1025 Commits

Author	SHA1	Message	Date
Tom Boucher	03b7125293	enhance(#3909): a probe that could not run no longer asserts a verdict (#3944) ... * test(#3909): failing-first suite for the fabricated probe fallbacks Binds the four fabrication sites found by executing the surfaces (ADR-3889 failure class (c)), each with a positive control so an over-firing fix goes red: - the blocking api-coverage.verify-pre gate certifying "no external-API integration" from a zero-byte phase scope - the assumption-delta query route scanning an unresolvable phase section as the empty string and reporting it as an examined negative - both capability fragments' probe fallbacks, which append a fabricated verdict rather than replacing, and fire on the legitimate exit-1 negative Verification runs on the remote runner. Refs #3909 * enhance(#3909): a probe that could not run no longer asserts a verdict ADR-3889 Phase 5. Four sites turned a failed or unexamined probe into a confident negative; each now reports what it could not establish. - check api-coverage.verify-pre: a phase with no plan body and no roadmap section ran detection over zero bytes and PASSED the blocking seal gate, certifying "no external-API integration" from input it never read. It now holds with scope_unavailable. The discriminator is bytes examined, never signals found, so a phase whose plans are real and simply carry no API vocabulary passes exactly as before. - query assumption-delta scan: an unresolvable phase section was scanned as the empty string and reported as an examined negative. It now returns {skipped, reason: phase_unresolved}, still at exit 0 — an ADR-2980 degraded result in the payload, leaving the gsd-tools exit projection to P8. - both capability fragments: `|| echo '{"detected":false}'` appended rather than replaced, and fired on the legitimate exit-1 negative, so a correct answer and an honest skip both arrived as two concatenated objects. They now keep the probe's own payload and manufacture only an explicit probe_unavailable skip when the probe produced nothing at all. Every registered outcome is more restrictive on a blocking gate, so this can turn a false green red and never a red green. Docs: FEATURES 156, CONFIGURATION (both keys), references/api-coverage.md seal-time outcome table, and a new how-to for the reason-code vocabulary. Verification runs on the remote runner. Closes #3909 * test(#3909): correct the stale unknown-phase assertion `unknown phase → detected:false, no throw (graceful)` scanned phase 999 against a two-phase roadmap and asserted `detected === false`. That pinned the fabrication as intended behavior: the phase does not exist, so the detector was handed the empty string and its "no core assumption changed" answer described nothing that was ever read. It now asserts the skipped-with-reason shape. The graceful-degradation contract the test was actually protecting — the query succeeds and does not throw on an unknown phase — is unchanged. Found by code review, not by the author. Refs #3909 * docs(#3909): author the FEATURES entry in its generator source `docs/FEATURES.md` is generated by `scripts/gen-features.cjs` from the per-feature fragments in `docs/features/`. The API-coverage entry was edited in the generated file, so the next regeneration silently dropped it. The text now lives in `docs/features/api-coverage-gate.md` and `docs/FEATURES.md` is regenerated from it, leaving the shipped file byte-identical and its content actually derivable. Caught by `lint:generated-sync`. Refs #3909 * test(#3909): bind the skip to "not found", and pin the discriminator The first verification run went red on one case, and the case was wrong rather than the code. `getRoadmapPhaseWithFallback` returns `null` for an unknown phase and for a missing ROADMAP.md, but for a section whose body is whitespace-only it returns the heading line alone — which is not empty. So a body-less section WAS found, and reporting `detected:false` over its heading is a real negative, not a fabrication. The test had assumed the resolver yielded `''` there. Correcting the test rather than the resolver keeps `skipped` bound to the distinction the issue asks for — found versus not found — and avoids diverging `assumption-delta scan` from `roadmap.get-phase`, which the fragment documents as sharing one resolver. Also adds the seeded property the test matrix had promised: for any plan body, the scope read back is whitespace-only exactly when the body was. That pins the gate's discriminator to bytes examined, so it cannot quietly become "no signals found", across unicode whitespace and CRLF. `docs/INVENTORY.md` picks up the reference doc's new seal-time outcome table — surfaced by the co-change gate, not by a lint failure. Refs #3909 * chore(#3909): backfill the changeset PR number Refs #3909 --------- Co-authored-by: sim <sim@local>	2026-08-27 15:50:12 -04:00
Tom Boucher	9410f7e6e6	enhance(#3897): ADR-3473 §8.3 rungs 2-4 — runtime marker, derived Codex sandbox, short-form depends_on (#3941) ... * test(#3897): failing-first coverage for §8.3 rungs 2-4 ADR-3473 §8.3 has four rungs; #3883/PR #3896 shipped the first. This pins the other three RED before any fix. Rung 2 — the install marker has four readers and resolveRuntime is not one. resolveRuntime resolves GSD_RUNTIME > config.runtime > 'claude' and reads no marker at all, while bin/install.js writes one (#2297) and FOUR hand-rolled readInstallRuntimeMarker copies exist: src/model-resolver.cts:65 (cached, with test seams), hooks/gsd-agent-isolation-guard.js:112, and TWICE in hooks/gsd-cursor-subagent-start.js at :346 and :355. Four copies of one rule. Fixtures and seam names mined from PR #3382 rather than re-derived; it implemented this rung and was closed "not on the merits". Rung 3 — the sandbox map, and the fallback that was the real defect. Measured across all 35 files in agents/, deriving workspace-write iff tools: declares Write or Edit: - all 11 CODEX_AGENT_SANDBOX entries derive to their mapped value exactly, zero disagreements — the map carries nothing the contract does not - 24 roles fall through `|| 'read-only'`, of which 16 declare Write or Edit So the map is redundant and the silent fallback is the defect. The maintainer chose to derive but hold those 16 at read-only pending the question of whether Codex enforces sandbox_mode or merely advises; HALT.md records it. T20 asserts the emitted sandbox_mode PER ROLE against a captured baseline, not in aggregate — an aggregate passes while one role silently widens, which is the proxy-instead-of-identity shape this repo names. T24 and T25 fail on a stale hold, so the hold list cannot rot into the subset map being deleted. Rung 4 — shortFormToId, recovered rather than invented. I nearly reported this as another wrong §8.3 claim: `git log -S shortFormToId` returns only documentation commits. That was the wrong instrument. Direct inspection of sdk/src/query/phase.ts at 11918dcc3^ shows five occurrences, and the tests match that code rather than a guess at its semantics — including first-write-wins on a duplicate short form. T43 asserts at the consumer's output: the emitted `waves` map from the real CLI, which pre-fix collapses to {"1":[...]} because every short-form edge is dropped. A unit assertion on resolveDependencyId would have passed throughout this defect's life. Observed RED, this tree: rung 2 11/11 fail — no marker rung, no seams rung 3 T23,T24,T25,T26,T30 fail; T28 fails (validate agents passes a TOML whose sandbox_mode disagrees — it checks presence only) rung 4 T42,T44 fail; T43,T49 fail with waves collapsed to a single wave 1 Green and staying green: T20/T21/T22/T27 as captured baselines, #3885's unresolvable-token warning and wave-verdict suppression, and #3785's display-mapping passthrough. If the third tier over-reaches, those go red — that is their job. Disclosed weakness: T45 (a canonical id with no dash is not short-form indexed) cannot be isolated behaviorally, because planMap always masks it. It is a non-crash boundary pin, weaker than the other rows, and is recorded as such rather than presented as equivalent. Design: .gsd/phase/feat-3897-adr3473-83-rungs/40-design.md Test matrix: .gsd/phase/feat-3897-adr3473-83-rungs/50-test-matrix.md Decision: .gsd/phase/feat-3897-adr3473-83-rungs/HALT.md Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * enhance(#3897): §8.3 rungs 2-4 — one marker reader, a derived sandbox, the third depends_on tier ADR-3473 §8.3 has four rungs. #3883/PR #3896 shipped the first. These are the other three. Rung 2 — the install marker had four readers, and resolveRuntime was not one. resolveRuntime resolved GSD_RUNTIME > config.runtime > 'claude' and read no marker, while bin/install.js writes one (#2297) and four hand-rolled readInstallRuntimeMarker copies existed: src/model-resolver.cts (cached, with seams), hooks/gsd-agent-isolation-guard.js, and twice in hooks/gsd-cursor-subagent-start.js. model-resolver's was already the house idiom, so it was promoted rather than replaced: src/runtime-slash.cts now owns it, and model-resolver plus both hooks delegate. The hooks reach it through ensureRuntimeBuild(), the seam lint-hooks-runtime-build-seam enforces. No import cycle existed - checked both directions before moving anything. The marker is the THIRD rung: env > project config > marker > 'claude'. N1 was checked rather than assumed, and my first reading of it was wrong. A marker holding an unknown name comes back essentially verbatim, which looked like a validation gap. Measured against the env rung with the same inputs - including "../../etc/passwd" and "claude;rm -rf /" - the two are identical, because they share resolveRuntimeNameFromCandidates. N1 asks for exactly that, and it is met. The residual (the shared normalizer normalizes shape, it does not validate against the known-runtime set) is pre-existing on the env rung and plausibly deliberate, since a new runtime should not need a code change. The marker also does not widen the trust boundary in any real sense: it lives inside the install tree beside the code, so anyone who can write it can write runtime-slash.cjs itself. Rung 3 — the map was redundant; the silent fallback was the defect. Measured across all 35 files in agents/, deriving workspace-write iff tools: declares Write or Edit: all 11 CODEX_AGENT_SANDBOX entries derive to their mapped value exactly, zero disagreements. The map carried nothing the contract did not already have, so it is DELETED rather than reconciled. What was actually broken is `|| 'read-only'`, which silently under-granted 24 of 35 roles. 16 of those 24 declare Write or Edit and would widen under derivation. Per the maintainer's decision (HALT.md), they are held at read-only pending the question of whether Codex enforces sandbox_mode or merely advises. Emitted TOML is therefore byte-identical for all 35 roles - asserted per role, not in aggregate, because an aggregate passes while one role silently widens. The hold list self-invalidates. A hold whose role no longer derives broader fails, and so does a hold naming a role with no agents/<name>.md. Without that it would rot into exactly the hand-maintained subset map being deleted, and this commit's own ledger claim would become false over time. Both cases were proved by injecting them and watching them throw. Two committed tests asserted the deleted map's existence and contents. They were pinning the thing being removed, so the tests moved rather than the production code: the 11 role-value pairs survive as a test-local PRE_3897_CODEX_AGENT_SANDBOX baseline, and the assertions now drive the real derivation against real agents/*.md. The coverage is preserved; only its source moved out of production code. validate agents gains checkCodexSandboxPosture, mirroring the existing checkCodexModelPosture: each installed TOML's sandbox_mode must equal the role's expected value, failing with role, expected and found. It previously checked file presence and manifest completeness only, so a TOML whose sandbox_mode disagreed passed. Rung 4 — shortFormToId, recovered rather than invented. I nearly reported this as another wrong §8.3 claim: git log -S returns only documentation commits. Wrong instrument. sdk/src/query/phase.ts at 11918dcc3^ carries five occurrences, and the implementation here matches that code rather than a guess at its semantics - including first-write-wins on a duplicate short form, deterministic from the sorted plan order. It resolves the bare plan number: depends_on: ["01"] now reaches 26-01-auth-hardening. That is a control-flow change, not a diagnostic one - plans that silently collapsed into a single wave 1 now execute in their declared waves, and execute-phase.md consumes those wave values. In-phase only, by construction: the map is built from this phase's rawPlans, so a same-named short form in another phase does not resolve. #3785's display-mapping passthrough and #3885's unresolvable-token warning and wave-verdict suppression are untouched and stay green. If the third tier had over-reached, those are what would have caught it. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3897): close a fail-open I introduced, and wire the posture check to its command Two blockers from review. Both are mine, and one is a security regression my own change created. 1. A held role could escape its hold by editing its own frontmatter. The Codex install loop set the sandbox identity from the agent's frontmatter `name:` field rather than from its filename, so the hold lookup keyed off a value the file itself declares: deriveCodexSandboxMode('gsd-doc-writer', <real file>) -> read-only deriveCodexSandboxMode('gsd-doc-writer-x', <same file, name: edited>) -> workspace-write deriveCodexSandboxMode('GSD-Doc-Writer', <same file, name: recased>) -> workspace-write What makes this a blocker rather than a nit is the DIRECTION. The deleted CODEX_AGENT_SANDBOX map had the identical lookup-key quirk, but it was an allowlist: an unmatched key fell back to read-only, which is safe. The new scheme derives workspace-write from the tool contract and uses the hold as a subtraction, so the same mismatch fails OPEN. I converted a fail-closed quirk into a fail-open one and did not notice; the isolated reviewer proved it by execution. Neither safety net caught it. validateCodexSandboxHolds only checks that <key>.md exists, never that a file's derived identity matches its key. checkCodexSandboxPosture looks the canonical source up by the installed TOML's filename, finds nothing for a renamed agent, and treats it as a custom non-roster agent — silently no violation. The identity is now the FILENAME STEM, which is what validateCodexSandboxHolds already validates and what an attacker editing frontmatter cannot change without renaming the file — at which point the existing validator catches it. The lookup is case-insensitive so a recase does not slip past either. The frontmatter name still drives the TOML body and filename, unchanged; only the sandbox identity moved. All 35 roster files were checked: name matches filename stem everywhere, so a stricter "they must agree or throw" invariant would have been safe against real content. It is deliberately NOT added — it would abort an install on a tampered file where emitting a correctly-derived read-only TOML is the safer outcome. Recorded as a fork rather than decided silently. 2. checkCodexSandboxPosture was exported and never called. cmdValidateAgents (src/verify.cts) called checkAgentsInstalled and checkCodexModelPosture only; grep for the sandbox check in that file returned nothing. So criterion 3 — "validate agents fails on semantic drift, not only on missing files" — was unmet, and `validate agents` behaved exactly as before. That is ADR-3473 Decision 2's named shape: a declared policy with no executor. It also meant the T28 test asserted at the helper's return value while the COMMAND stayed broken — the ADR-3180 Decision 4(b) failure this epic exists to close, committed by me while enforcing it elsewhere in the same epic. Now wired as an additive `sandbox_posture` field beside `codex_posture`, following the sibling precedent exactly. Drift is report-only, not a non-zero exit, because that is what checkCodexModelPosture does — two sibling posture checks disagreeing about whether a violation is fatal would be its own defect. The choice is recorded in a comment rather than left implicit. A consumer-output test now drives the real CLI and asserts on the emitted JSON, and was shown failing before the wiring and passing after. Also corrected a stale artifact: the design's Known limit L1 still claimed rung 3 was not in this deliverable, written while it was halted and false once the maintainer unblocked it. Verified after both fixes: the three bypass probes all return read-only, the per-role table is 35/35 byte-identical, and both hold self-invalidation cases still throw. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(#3897): the marker rung, the derived sandbox, and the bare plan-number depends_on Reference: the runtime precedence ladder in docs/CLI-TOOLS.md gains the install marker rung; docs/COMMANDS.md documents validate agents' new sandbox_posture field; docs/reference/plan-md.md documents that depends_on accepts the bare plan number. Explanation: a docs/features fragment keyed id 3897, so it cannot collide with a concurrent PR hand-allocating a section number, regenerated into FEATURES.md. ADR-3473 §8.3 gains an ANSWER blockquote in the document's own correction style, recording what was measured and built against the section's 2026-08-26 correction - including the qualification that checkAgentsInstalled itself still checks presence only, and the semantic assertion lives in a sibling wired into validate agents rather than folded into it. No how-to. Both user-visible changes are zero-step: a non-Claude install resolving its own runtime, and plans executing in their declared waves, both happen without the user doing anything. docs/how-to/control-the-reported-host-runtime.md covers a DIFFERENT ladder (resolveReportedRuntime / agent_runtime) that this change does not touch, and was deliberately left alone rather than edited by association. No tutorial - nothing multi-step to walk through. docs/AGENTS.md unchanged: it documents Claude-side tools frontmatter, never Codex sandbox_mode, and the emitted tools contract did not change. The prompt layer documents depends_on only by example, not by schema, so nothing there needed editing - and few-shot-examples/plan-checker.md already showed depends_on: ['01'], which now actually resolves. Translated copies of plan-md.md are untouched; the project treats translations as community-maintained. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3897): move the sandbox derivation out of the installer, off the install path, and off a third parser The full suite came back with 26 failures across four files. Three distinct causes, mapped individually rather than assuming the first explained the rest. A. Requiring bin/install.js printed the GSD banner to stdout and corrupted `validate agents` JSON. Unexpected token '', "[36m ██"... is not valid JSON checkCodexSandboxPosture reached deriveCodexSandboxMode by lazily requiring bin/install.js, whose module load prints the ASCII banner. So the command emitted banner bytes before its JSON and every JSON consumer broke, including ten tests that predate this branch. src/ reaching into bin/ was backwards layering that happened to also be loud. The derivation now lives in src/codex-agent-toml.cts - the existing Codex TOML domain module, no new module and no six-gate ripple - and both bin/install.js and src/agent-install-check.cts import it. One owner, which is §8.3's rule applied to the fix for §8.3. B. The stale-hold throw fired on a legitimate partial source dir, and masked a security assertion. validateCodexSandboxHolds treated "this hold's .md is absent from the install SOURCE dir" as a stale hold and threw. A test fixture, or any partial install source, legitimately contains a couple of agents. Worse, it threw BEFORE the path-escape check, so a test asserting that a `../../evil` frontmatter name is rejected got my unrelated error instead of the traversal rejection it was written for. A fail-closed check of mine was hiding a real security check. The "no stale holds, shrink-only" invariant is a property of the repo's canonical agents/ roster, not of whatever directory an install happens to read. It is off the runtime path and enforced where it belongs, in the tests that already existed for it. A partial source dir now installs cleanly, and the evil-name case throws with its own escapes-configHome message again. C. T8 depended on ambient process.env state. The marker/env parity assertion round-tripped through live process.env. It now compares against resolveExplicitRuntime's already-exported dependency-injection parameter - deterministic and hermetic, same claim. Proven still falsifiable rather than assumed: with the marker rung's normalization temporarily bypassed the two rungs diverge ("codex\n../../etc/passwd" vs "codex-../../etc/passwd") and the assertion fails, then passes again once reverted. One correction folded in along the way. The first version of the move added private _extractFrontmatterAndBody/_extractFrontmatterField helpers to codex-agent-toml.cts - a THIRD copy of frontmatter extraction, where the graph already shows two (bin/install.js:2348, runtime-artifact-conversion.cts:893). Adding a third inside the epic whose thesis is one implementation per rule is not defensible. deriveCodexSandboxMode no longer parses anything: it takes (identity, toolsValue) and each caller supplies the tools value using the extractor it already has. Both helpers are deleted. The identity argument is still the filename stem, so the fail-open fix is untouched. Verified after all three: `validate agents --raw` emits parseable JSON with no banner and both posture fields; the four hold-bypass probes still return read-only; the per-role table is 35/35 byte-identical at 26 read-only / 9 workspace-write; the hold list is still 16. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3897): drop a dev-only transitive dep, make the derivation total, retire a stale fallback test Suite down to 7 failures from 26. Three more causes, mapped individually. A. My extractor import dragged in a script that does not exist in an installed tree. Cannot find module '../../../scripts/fix-slash-commands.cjs' Chain: src/agent-install-check.cts imported runtime-artifact-conversion.cjs, which requires command-roster.cjs, whose line 36 requires ../../../scripts/fix-slash-commands.cjs. That path exists in the repo and not in an install, so every test exercising a synthetic install dir died at module load. I picked that extractor for convenience without checking what it pulls in - the same mistake that produced the banner bug, one layer further out. agent-install-check now uses a single-purpose extractToolsLine on codex-agent-toml.cts. That is deliberately NOT a general frontmatter parser: we deleted those helpers a commit ago for good reason, and this reads one line. Verified from outside the repo root that requiring either module prints nothing and does not throw. B. A test pinned the deleted name-based fallback. 'defaults unknown agents to read-only' called generateCodexAgentToml with a fixture declaring tools: Read, Write, Edit. Under derivation an unknown agent with a writing contract correctly derives workspace-write - design row S6, a new writing role gets the contract, not the pin. The behavior it asserted was the silent fallback this rung deleted; identity no longer decides the sandbox. Replaced with two rows rather than a flipped string: no tools declared -> read-only (absence is not a grant), and Write/Edit declared -> workspace-write. Strictly more coverage than the row it replaces. C. The stale-hold check still threw per derivation call. Last commit took the roster-existence check off the install path, but deriveCodexSandboxMode itself still threw when a hold's role did not derive broader FOR THE CONTENT IT WAS HANDED - so it fired on any synthetic fixture for a held role. The throw is gone, and it cost nothing: if a held role's content does not derive broader, the hold pins read-only and derivation returns read-only anyway, so the hold is a no-op and there is nothing to fail about. The staleness invariant is a property of the real agents/ roster, and validateCodexSandboxHolds still enforces it there - confirmed against the real roster after the change, not assumed. deriveCodexSandboxMode is now total: every (identity, toolsValue) including undefined and null returns read-only or workspace-write, never throws. Verified: validate agents emits parseable JSON; the four hold-bypass probes return read-only; the per-role table is 35/35 at 26 read-only / 9 workspace-write. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(#3897): put the rung-3 decision in the shipped docs instead of pointing at an ignored path The ADR entry and the feature fragment both ended their rung-3 explanation with "see .gsd/phase/feat-3897-adr3473-83-rungs/45-decision-rung3-sandbox.md". That directory is gitignored (.gitignore:55), so the rationale for holding 16 roles at read-only was reachable only from the machine that produced it. A reader of the ADR got a pointer to nothing. Both now carry the reasoning inline: the criterion asks both that the sandbox derive from the declared tool contract and that no role gain a broader sandbox, and those cannot both hold, because a faithful derivation widens 16 roles the deleted map never listed and that fell through its silent read-only default. The resolution is derive-and-hold - the derivation owns the rule now, each hold is released as its enforcement question is answered, and a hold is reversible where a widened sandbox that turns out to be enforced is not. Checked before assuming this was a defect class: CONTEXT.md cites .gsd/phase/<slug>/40-design.md as its standard Design: provenance line in eight module entries, and four other shipped docs do the same. Citing a phase artifact is an established convention here, so those are left alone. What was wrong was specific to these two: they put load-bearing rationale behind the pointer instead of provenance. docs/FEATURES.md regenerated from the fragment via scripts/gen-features.cjs rather than hand-edited. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3897): close a fail-open, stop a silent mis-resolution, and read a declaration as a declaration Two orthogonal reviews on the shipped sha. Three of the findings are the same failure class this epic exists to close, committed inside it. 1. BLOCKER - the sandbox was decided for one identity and applied to another. bin/install.js derived sandbox_mode for the filename stem and then wrote the result to `${name}.toml`, where name comes from the file's own frontmatter. Make the two disagree and a HELD role's artifact goes wide: rename gsd-doc-writer.md -> gsd-doc-writer-v2.md, keep name: gsd-doc-writer -> stem is unheld, derives workspace-write, lands on gsd-doc-writer.toml add any gsd-*.md whose frontmatter name: is a held role -> clobbers that role's toml with workspace-write Both emit read-only on origin/next, because the deleted map was an allowlist and a miss fell back safe. This is a regression my change introduced. The previous review round moved the HOLD KEY off frontmatter to the filename stem and left the OUTPUT PATH on frontmatter; my own comment at install.js:6985 calls that value attacker-editable, four lines above the line that uses it as the filename. The decision is now made over BOTH candidate identities, most-restrictive wins: if either the stem or the emitted name is held, the mode is read-only. 2. MAJOR - hold matching was toLowerCase() only, so confusables escaped. Turkish dotted/dotless i, fullwidth, NFD, trailing space/NBSP/dot/newline, ./ and ../agents/ all slipped the hold and emitted workspace-write. Identities are now basenamed, trimmed of NBSP/zero-width/control characters, NFKC-normalized and lowercased - and anything still carrying a character outside [a-z0-9._-] is treated as suspicious and derives read-only. We do not enumerate confusables; every shipped roster file is ASCII, so refusing to widen on an identity we cannot recognize is fail-closed with no false positives on real content. 3. MAJOR - the short-form depends_on tier mis-resolved SILENTLY. shortFormToId keyed on the last dash-segment of any canonical id with no constraint that it is a plan number, so a phase holding 09-FIX-auth-PLAN.md made depends_on: ["auth"] bind at wave 2 with zero warnings. This is the worst shape in the epic: the unresolvable-token warning fires on a DROPPED token, so a MIS-RESOLVED one is invisible and the tool reports a confident wave assignment built from a wrong edge. A wrong edge is worse than a missing one. The segment must now match /^\d+$/, which is exactly the contract docs/reference/plan-md.md already documents. This tier was recovered verbatim from the retired SDK lineage, which carried the same defect; we are deliberately NOT preserving it bug-for-bug, and the comment says so, so the next reader does not "restore" it. 4. MAJOR - the derivation was reading a declaration as an absence. extractToolsLine read one line, so a YAML list-form tools: block returned only its first item. Two roster files use list form, and gsd-nyquist-auditor declares Write and Edit there - parsed as "- Read", found no write tool, and emitted read-only. Rung 3's headline claim is that sandbox_mode derives from the declared tool contract; that claim was false for 2 of 35 roles and materially wrong for 1. Reading a declaration as an absence is the silent-drop class this epic exists to close. Renamed extractToolsValue and taught it both shapes. gsd-nyquist-auditor now derives workspace-write and joins CODEX_SANDBOX_HOLDS as its 17th entry, per the standing derive-and-hold decision - so emitted TOML stays byte-identical at 26 read-only / 9 workspace-write while the hold list finally records every role that would widen. A previous pass declined this fix because it moved the count; that inverts the priority. Byte-identity is preserved THROUGH the hold, not by leaving a parser broken. Divergence check, because this is where that bug hides: both paths feeding sandbox derivation - install.js's emitter and checkCodexSandboxPosture - now route through the one extractor. The tools readers in runtime-artifact-conversion and install.js's other frontmatter call sites serve Claude-side emission and do not feed sandbox derivation. Also fixed, each real: the posture check's `found` used a naive whole-file regex where its own sibling uses the block-aware scanner, so prose inside developer_instructions produced a false violation; `found` skipped truncatePostureValue and leaked a 300-char value into validate agents output; deriveCodexSandboxMode's absolute never-throws claim was false for an object with a throwing toString; T49 could not falsify cross-phase leakage (its target phase had its own 01, so a globally-scoped map passed too); T20/N6 iterated a hardcoded table and pinned the FIXTURE size, so a 36th agent would be silently unchecked; three tests reimplemented the code they were testing instead of importing it; and T2-T4 deleted GSD_RUNTIME without restoring it. Verified: hold list 17, gsd-nyquist-auditor derives workspace-write unheld and emits read-only held, roster 35/35 at 26/9, depends_on ["auth"] no longer resolves while ["01"] still does, both identity-bypass cases and every confusable vector emit read-only. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(#3897): the hold list is 17, and the reason the 17th was missing The count read 16 because the derivation could not read the declaration it claimed to derive from: the tools reader was single-line, so a YAML list-form tools: block returned only its first item and gsd-nyquist-auditor's declared Write and Edit were read as an absence. Both the ADR entry and the feature fragment now carry the corrected count and the reason for it, rather than a silently updated number. Deriving from a declaration you cannot parse is not deriving, and a flattering count is worse than a wrong one because it looks settled. docs/FEATURES.md regenerated from the fragment. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#3897): backfill changeset pr number Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-27 15:19:01 -04:00
Tom Boucher	34399eed70	docs(#3942): ADR — the emitted-drift ack belongs in a commit trailer (#3943) ... The acknowledgment explains one PR's unattributable emitted-artifact delta. The moment that PR merges the delta is in the base, so the acknowledgment can never clear anything again — its useful life is exactly the PR's open window. It is stored in permanent, shared, merge-path state. That single lifetime mismatch is the origin of six consecutive rounds of defect-and-fix (#2789, #2914, #3078, #3823, #3842, #3875), each fix generating the next defect, ending in a scheduled sweeper whose own PR (#3927) could not merge itself. ADR-3942 supersedes ADR-2719 section 3 only, and its #2789 Amendment. Sections 1, 2 and 4-7 are retained and depended upon: the conservation law itself is not in question, only where its escape hatch is stored. Also lands the root-cause research note behind the decision, and regenerates the ADR index. Two corrections to earlier framing, recorded here rather than dropped: - Per-commit trailers do NOT reliably survive squash-merge; the repo's own ship.md:306 reconstructs a gate trail precisely because squash discards it. The ADR therefore claims no durable audit record on next. Non-survival is the property being bought, not a cost. - The test job checks out at depth 1 (test.yml:107-110, no fetch-depth key). A depth-1 checkout cannot see the PR's commit range and fails VACUOUSLY rather than erroring, so fetch-depth: 0 is a named decision in the ADR and the acceptance criterion requires a test that fails under depth 1. Docs-only; no code, no changeset. Refs #3942 Co-authored-by: sim <sim@local>	2026-08-27 14:06:29 -04:00
Tom Boucher	1e67ec9737	enhance(#3908): the scanners distinguish an empty diff from one they could not compute (#3937) ... * feat(#3908): the scanners distinguish an empty diff from one they could not compute collect_files ended 2>/dev/null || true, which destroyed the evidence three ways: the redirect discarded git's diagnostic, the pipe replaced git's status with grep's, and || true forced success regardless. Four distinct conditions - an established-empty diff, a bad ref, no repository, and a repository with no commits - all reported clean, and a secret scanner reporting clean because git failed is indistinguishable from an all-clear to any gate consuming it. git now runs separately from the filter so its status and diagnostic both survive. An established-empty diff exits NO_INPUT; a scope that could not be established exits UNAVAILABLE; the usage sites move off 2 to USAGE. || true is retained on the filter alone, where it is correct: a diff of only images is empty, not failed. Codes are sourced from a generated shell fragment rather than written into three scripts, so a re-allocation cannot desync them, and a missing fragment fails loudly instead of falling back to literals. The security workflow is updated in the same change: without it, a docs-only PR would newly fail the job. * fix(#3908): keep scanner stderr out of the file list, and drop try/finally from test bodies Capturing git and find output with 2>&1 was right for the failure path but wrong for the success path: a warning emitted alongside a successful diff flowed into the file list and was treated as a filename. stderr is now captured separately, forwarded as a warning on success and as the diagnostic on failure, and never folded into the list. Also converts the control tests' try/finally blocks to t.after(), which CONTRIBUTING bans inside a test body because it masks failures. * chore(#3908): backfill changeset pr number * docs(#3908): record the scanners' four-outcome exit contract SECURITY.md is root-level, so the docs gate correctly held: a Changed fragment owes a file under docs/. The contract also belongs where the feature is described, as REQ-SCAN-INJ-05. docs/FEATURES.md is GENERATED from per-feature fragments (#3840) - the first edit went into the generated file and gen-features --check caught it, which is the same edit-the-output drift this epic exists to close. The fragment is the source; FEATURES.md is regenerated. --------- Co-authored-by: sim <sim@local>	2026-08-27 13:11:13 -04:00
Tom Boucher	c5f2b94b27	enhance(#3907): gates report no-input instead of a verdict they never reached (#3932) ... * feat(#3907): gates report no-input instead of asserting a verdict they never reached The three stdin-reading gates bound 2 to a stdin read error only, with no arm for stdin closed at zero bytes - so empty input flowed into the detector, found nothing, and exited 1, which each module's own comment defines as a negative verdict. An unset PHASE_SECTION made the UI gate assert the phase has no UI. Empty and whitespace-only input now exit NO_INPUT, and a read error exits UNAVAILABLE rather than a locally-invented 2, both resolved through the registry and delivered by terminateNow. The exit code was only half of it: under --json the same input emitted {detected:false}, byte-identical to the fabricated payload #3909 exists to fix, and the blocking coverage gate reads that payload. Empty input now emits the in-tree {skipped:true,reason} form with no detected key at all. teams-status is excluded: it never reads stdin and has no invented 2, so the four-module framing in the issue and ADR is wrong. The dead root bin/lib/ui-safety-gate.cjs is deleted - no installer reference, no workflow invocation, and the live fallback chains are for other modules. Its removal restores the unit tests to the module that actually ships; they had been asserting the stale copy's two-field shape, which is why it drifted unnoticed. * fix(#3907): drive gate tests through the process seam, and make removed-but-needed basename-precise CONTRIBUTING requires every subprocess go through tests/helpers/process-seam.cjs; two of the three gate suites hand-rolled spawnSync while the third, added in the same change, used runNode correctly for the identical injection case. Converted the blocks this change added, leaving pre-existing ones alone. Deleting one of two files sharing a basename made lint-removed-but-needed report 14 references that were all to the surviving canonical module - the false-positive class its own docstring names. It now matches on the deleted file's full path when a surviving file shares its basename, which is more precise rather than weaker: a genuine full-path reference still fails, and behaviour is unchanged when no basename collides. It immediately caught a docstring on this branch that spelled the deleted path. * test(#3907): update the one existing assertion that pinned the old empty-stdin verdict A pre-existing test asserted exit 1 on empty stdin - the defect this phase removes - and was missed because the change added new blocks without auditing existing ones pinning the old contract. Audited the rest: the other three status-1 assertions in that file all feed real input and are the genuine-negative controls that must keep returning 1, so exactly one was stale. The retired 2 is gone from the describe's contract comment too. * chore(#3907): backfill changeset pr number --------- Co-authored-by: sim <sim@local>	2026-08-27 11:37:31 -04:00
Tom Boucher	929e02cb2c	enhance(#3885): no silent swallow, and no verdict manufactured from dropped data (#3925) ... * test(#3885): failing-first coverage for the depth bound and the manufactured wave verdict ADR-3473 §8.5 says a swallowed failure may not become an authoritative-looking answer. Three families do exactly that today; this commit pins each one RED. Measured on this tree, 2026-08-27: intel query, .planning/intel/file-roles.json nested 12000 deep -> exit 1, "Error: Maximum call stack size exceeded" searchJsonEntries / matchesInValue carry no depth parameter at all. The MAX_JSON_SEARCH_DEPTH = 48 bound existed in the retired SDK lineage (sdk/src/query/intel.ts at 11918dcc3^) and the surviving .cts lineage never received it. same fixture nested 48 and 49 deep -> both return total=1 at exit 0, truncated=undefined Nothing distinguishes "searched to the bottom" from "stopped looking". query phase-plan-index, a plan whose depends_on names an unresolvable token -> warnings: ["Plan 03-02: declared wave: 2 but depends_on DAG places it in wave 1"] The token is never mentioned. computeDependencyLevels drops the edge with `if (!resolvedDep) continue;`, every plan becomes a root, and the tool then reports the author's correct wave: as the thing that is wrong. countPhasePlansAndSummaries with fs.readdirSync throwing EACCES -> hasContext:false, indistinguishable from a phase that simply has no CONTEXT.md. context_read_error is undefined. The shapes these tests assert against, chosen here so the implementation has a target rather than inventing one later: `truncated: boolean` on the intel query result, `unresolved: Array<{plan, token}>` from computeDependencyLevels, and `context_read_error: string | null` per analyzed phase. Deliberately green, and they must stay that way — each stops the fix from over-firing: depth 48 is found and NOT flagged truncated (the ceiling is inclusive) a shallow miss reports no truncation (noise control, N1) 10,000 siblings at depth 2 are unaffected (the bound is DEPTH, N2) a genuine wave: mismatch on a fully-resolved DAG still warns (N3) a genuinely missing directory is absent, not an error the emitted depends_on display mapping still passes an unresolved token through verbatim — already pinned by the existing #3785 test, so no duplicate was added T31 asserts at the consumer's output per ADR-3180 Decision 4(b): it runs the real CLI and reads the emitted JSON, because a unit assertion on computeDependencyLevels would have passed throughout #3427's life. Design: .gsd/phase/feat-3885-no-silent-swallow/40-design.md Test matrix: .gsd/phase/feat-3885-no-silent-swallow/50-test-matrix.md Refs #3885 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * enhance(#3885): no silent swallow, and no verdict manufactured from dropped data Implements ADR-3473 §8.5. A failure or a gap in the input stops being absorbed into an output that reads as authoritative. The recursion bound, restored but NOT verbatim (src/intel.cts) MAX_JSON_SEARCH_DEPTH = 48 is threaded through searchJsonEntries and matchesInValue, which carried no depth parameter at all. The bound existed in the retired SDK lineage (sdk/src/query/intel.ts at 11918dcc3^) and the surviving .cts lineage never received it — §8.3's "a consolidation may not delete an invariant along with the surface that held it", demonstrated. Measured before: a .planning/intel file nested 12000 deep exits 1 with "Error: Maximum call stack size exceeded". Reachable from a project document. The original returned a bare `false` at the ceiling. Restoring that verbatim would trade a crash for a silent "no match" when the truth is "I stopped looking" — the same class this epic exists to close, and ADR-3473 Decision 4 forbids it. So the bound carries a truncation signal: nesting 47 -> found, truncated false nesting 48 -> found, truncated false (the ceiling is inclusive) nesting 49 -> not found, truncated TRUE nesting 12000 -> exit 0, truncated TRUE, no RangeError A shallow document that simply has no match reports truncated FALSE — the flag means "I stopped early", never "I found nothing", or it would be noise. The bound is on DEPTH: 10,000 siblings at depth 2 are unaffected. The dropped edge is named, and stops being blamed on the author (src/phase.cts) computeDependencyLevels dropped every unresolvable depends_on token with a bare `continue`. Each drop makes a plan a root, so the whole phase collapses to wave 1 — and cmdPhasePlanIndex then reported the author's CORRECT wave: as the thing that was wrong. Before: warnings: ["Plan 03-02: declared wave: 2 but depends_on DAG places it in wave 1"] After: warnings: ["Plan 03-02: depends_on token \"nonexistent-token-3427\" does not resolve to any plan in this phase — edge dropped, wave placement for this plan may be unreliable"] The suppression is PER PLAN, never blanket: a plan with a fully-resolved DAG and a genuinely wrong wave: still gets the mismatch warning. resolveDependencyId stays two-tier — the shortFormToId third tier is §8.3/Phase 6's rule and is deliberately not built here. The emitted depends_on display mapping still passes an unresolved token through verbatim (#3785). No artifact from failed inputs (gsd-core/workflows/review.md, #3352) A failed lane leaves no result file, so "every lane failed" is exactly "the aggregate JSONL has zero lines" — the gate condition already existed as a byproduct. REVIEWS.md is no longer written in that case, and the commit step is skipped with it. A budget-SKIPPED lane also leaves no file and is NOT counted as a failure. Per-lane output and non-empty .err are preserved to .review-diagnostics/ before `rm -rf "{run_dir}"` destroys the only record that the lanes failed at all; the commit step names one file, never a glob, so the diagnostics are not swept in. Unreadable is not absent (roadmap.cts, gap-checker.cts, init.cts x2) Four callers collapsed an EACCES on a phase directory into [] and reported hasContext:false — byte-identical to a phase that simply has no CONTEXT.md. Each now names the directory it could not read. A genuinely missing directory stays absent rather than becoming an error, which is what keeps the fix from over-firing. Fatal errno folded into a retry set: audited, no defect found Reported as a verified negative rather than padded with a change. withPlanningLock was fixed by #1884/PR #3472; acquireStateLock by #3776; atomicRenameWithRetry and estimate-cli's renameWithRetry are correct by construction — bounded set {EPERM,EBUSY,EACCES}, bounded attempts, and they return or rethrow the final error rather than swallowing it. estimate-cli's sole caller surfaces that rethrow as write_error in its JSON output. Manufacturing a diff to make the checkbox look worked-on is the Goodhart outcome Decision 6 exists to prevent. Disclosed: R46 (the commit step names one file, never a glob) is a real regression guard but is NOT independently failing-first — the commit fence is byte-identical pre- and post-fix, so it only fails pre-fix through its shared extraction dependency. Recorded rather than claimed as fail-first. Design: .gsd/phase/feat-3885-no-silent-swallow/40-design.md Test matrix: .gsd/phase/feat-3885-no-silent-swallow/50-test-matrix.md Refs #3885 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3885): escape untrusted tokens, and stop cleanup destroying unpreserved evidence Two review findings, both real, both in my own change. An isolated adversarial review found the evidence-preservation block never checked mkdir/cp exit status while `rm -rf "{run_dir}"` ran unconditionally in a SEPARATE fenced block. A disk-full or unwritable phase directory therefore still destroyed the only copy of the failed lanes' output — reintroducing the exact #3352 data loss this item exists to stop, inside the fix for it. Preservation and cleanup are now one block, because each fenced block is a separate execution and a shell variable cannot carry between them. mkdir -p and each cp are exit-checked; cleanup runs only when preservation succeeded, and a failure warns naming the intact run directory. "Nothing to preserve" is not a failure and still cleans up. Driven three ways: success removes run_dir, failure leaves it intact with the warning, nothing-to-preserve removes it. The failure is induced by a file-vs-directory conflict rather than chmod 0o000, which root bypasses. The new unresolved-depends_on warning embedded a user-authored token verbatim: warnings: ["Plan 03-02: depends_on token \"evil Plan 03-01: FORGED WARNING\" does not resolve ..."] The JSON wire form is safe, and the security reviewer judged it non-exploitable for that reason. It is escaped anyway through formatDiagnosticToken — the helper #3884 added one phase earlier for exactly this class. warnings[] is an array a consumer naturally prints line by line, and not reusing the sibling fix is the generative-fix-divergence shape this epic exists to close. The same treatment is applied to context_read_error / phase_dir_read_error, which embed a phase directory path a repository can choose, and to the fs error message, which echoes the raw path itself. Known limit L5 recorded: the bound is on DEPTH only. A 300,000-element shallow array yields a 14.5MB reply with truncated:false. Correct per §8.5 and per negative space N2, disclosed rather than left to be discovered. Refs #3885 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3885): unreadable is not absent in intel.cts either, and a corrupt snapshot is not "no snapshot" Blocker from the round-2 isolated review, and it is my own inconsistency: this phase applied "unreadable is not absent" to phase directories and left it broken in the file it was already editing. chmod 000 .planning/intel/file-roles.json gsd-tools intel query <term> -> {"matches":[],"total":0,"truncated":false} exit 0 safeReadJson swallowed every read failure and returned null, so an EACCES was byte-indistinguishable from an absent file AND from a genuine no-match. Now it separates three states: ENOENT stays silently absent, because not every project has every intel file and intelQuery loops over all of them expecting misses; EACCES/EIO and malformed JSON are both surfaced naming the file. A corrupt intel file previously read as "no matches" too — same defect, same fix. Threading that outcome through the other three callers found something worse than the reported case. intelDiff returned no_baseline:true for a corrupt or unreadable snapshot — not a silent failure but an actively FALSE verdict, telling the caller they never took a snapshot when they did. That is §8.5's headline case, so it is fixed and tested rather than noted. intelStatus and intelApiSurface collapsed the same way; intelApiSurface additionally printed a "not yet populated" banner that was simply untrue. Every row is failing-first, including the absent-file ones — the field is new, so it does not exist pre-fix at all. Those rows are not pre-fix pins; they pin that the fix does not OVER-fire on the ordinary absent case, which is what would turn this into noise on every project lacking an intel file. IO failure is injected by monkeypatching fs and restoring in finally, never chmod 0o000 — root bypasses mode bits, so the reviewer's manual chmod repro is not reproducible as a test. Refs #3885 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test(#3885): build the pathological intel fixture as text, not by stringifying a nested object The remote runner came back red on Linux with two failures, both T4: deeplyNestedIntelDoesNotOverflowTheStack, while the same test passed on macOS. The product was never at fault. writeNestedFixture(12000) built a 12,000-deep JavaScript OBJECT and then JSON.stringify'd it. JSON.stringify recurses once per level, so it overflowed the TEST PROCESS's stack — the error was thrown before the CLI was ever spawned. Linux's container stack is smaller than macOS's, which is the whole of the platform difference. Measured, with the same document built as JSON TEXT so nothing in the building process recurses: depth=100 rc=0 truncated=true depth=5000 rc=0 truncated=true depth=12000 rc=0 truncated=true depth=60000 rc=0 truncated=true V8 parses this shape iteratively; only stringify recurses. The bound works at every depth tried. The fixture is now built by string concatenation. That is also the more faithful input — a real deeply nested JSON document on disk is exactly what the bound guards, where a stringified object was only ever a way to produce one. The depth stays 12000. Lowering it would have made the test pass by weakening it to accommodate a fixture bug, and 12000 is a legitimate pathological input the product handles. T4 remains a genuine fail-first: rebuilt against the parent of the commit that added the bound, the string-built depth-12000 fixture still drives the CLI to rc=1 with "Error: Maximum call stack size exceeded". A comment records why the fixture is text, so it is not "simplified" back into a macOS-green / Linux-red test. Refs #3885 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#3885): backfill the changeset PR number Refs #3885 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test(#3885): normalize path separators before splicing into the workflow's bash CI red on one lane — test (windows-latest, 24, shard 3/3). macOS, Linux and the remote runner were all green. AssertionError: commit must name the single REVIEWS.md file; got: --files C:UsersRUNNER~1AppDataLocalTempgsd-3352-phasedir-mOKmuy/03-REVIEWS.md Every backslash in C:\Users\RUNNER~1\AppData\Local\Temp\... was eaten. The harness spliced an OS-native temp path into the extracted bash, and bash consumes \U, \A, \L and \T as escapes on an unquoted expansion. The same loss broke RUN_DIR, so "rm -rf" targeted a path that never existed and the run directory survived — which is the other two assertions. This is a fixture defect, not a product one, and that was checked rather than assumed. In production the phase directory is toPosixPath-normalized at every call site that serializes it (bin/lib/init.cjs:951, 1381, 1461, 1529, 1595), and the run directory is created by "mktemp -d" running inside the bash block itself (gsd-core/workflows/review.md:163), which emits POSIX-style output even under Git-Bash on Windows. Neither ever carries a backslash where the workflow reads it. The file's pre-existing #3034 harness splices raw native paths too, but only ever inside double-quoted assignments, so it never tripped this — my new harness followed that convention faithfully into the one place where it does not hold. Both now splice through toPosixPath from shell-command-projection, the established seam, which is a no-op on POSIX and mirrors what production does. No assertion was weakened. "commit must name the single REVIEWS.md file" and "the run dir must still be destroyed" still assert exactly that; only how the fixture supplies its path changed. Nothing is skipped on Windows — a t.skip() here would have hidden the question of whether the exposure was real, which is the question that mattered. Driven both ways: a synthetic C:\Users\RUNNER~1\... input reproduces the exact CI string when unfixed and yields C:/Users/RUNNER~1/... when fixed; a POSIX input produces a byte-identical shape, proving the normalization is idempotent. Refs #3885 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test(#3885): stop the harness making the deleted run dir its own cwd Windows shard 3/3 stayed red after the separator fix, on two assertions the separator fix never touched: AssertionError: the run dir must still be destroyed AssertionError: nothing to preserve is not a failure — run dir must still be removed The separators were a real bug and fixing them fixed the --files assertion. They were not this bug, and two CI cycles went into the wrong axis before I stopped converting path forms and looked at what the harness actually does. runWriteReviewsFlow passed cwd: runDir to runHook, so the child bash process's working directory WAS the directory the block under test then removes with rm -rf "$RUN_DIR". POSIX allows a process to delete its own cwd — verified locally, cd "$d"; rm -rf "$d" removes it cleanly — and Windows does not: a live process's working directory cannot be removed. So on Windows the directory survived and both assertions failed, on macOS and Linux it vanished and they passed. Nothing to do with slashes. Harness-only. Production never cd's into the run directory; every reference is by absolute path, and RUN_DIR is created by mktemp -d inside the bash block itself (gsd-core/workflows/review.md:165) rather than injected. review.md is unchanged. Fix: the child now runs with its cwd in an unrelated temp directory that the block under test never deletes. Neither assertion was weakened, and nothing is skipped on Windows — the tests in this file carry no platform guard and run there unconditionally, which is how this surfaced at all. Honest limit: the Windows failure mode cannot be reproduced on macOS, because POSIX permits the very thing Windows refuses. The diagnosis is grounded in that documented divergence and in the fact that only the Windows lane failed, but the green outcome on windows-latest is unverified until CI runs it. Refs #3885 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-27 04:12:47 -04:00
Tom Boucher	941b62249e	enhance(#3906): two terminators over one registry, with a versioned exit projection (#3924) ... * feat(#3906): two terminators over one registry, with a versioned projection Adds terminateNow (write-then-terminate, for callers that cannot wait for the event loop) beside runMain (drain-then-exit), both projecting through one shared function so they cannot disagree - the parity the ADR makes mandatory. A failed write does not change the exit code: letting it propagate would fail a hook open, which is what the fail-closed branches exist to prevent. The projection is versioned. v1 reproduces today's integers, including keeping a payload-carried degraded result at exit 0 - ADR-2980 ratified that across 60 sites and declined normalizing it on measured blast radius. v2 applies the registry. --exit-contract=v2 or GSD_EXIT_CONTRACT=v2 selects it; an unrecognized version throws rather than silently defaulting. The registry is now emitted beside both copies of the exit module, so it resolves as a sibling in the built tree and in the committed scripts/ copy that must load on an unbuilt clone. * fix(#3906): actually restrict code 2 to terminateNow, and generate the registry's type The claim that terminateNow is the only place 2 can be produced was false: runMain's outcome arm applied no guard, so runMain(()=>'HOOK_DENY') set exitCode 2 through the drain path - and the parity matrix demonstrated it while calling it parity. runMain now refuses any outcome projecting to the hook-protocol code, gated on the code rather than the name so an alias cannot slip past, and the matrix asserts the restriction instead of contradicting it. The ambient type for the generated registry was hand-written with no gate against the generator's actual output - the declared-surface-diverges-from-runtime defect class this epic exists to close, reintroduced inside it. It is now a third generated artifact covered by the same --check. Also converts every test-body try/finally to t.after(). * test(#3906): derive the glossary fixture's dependencies instead of hand-listing them Adding a require to scripts/lib/cli-exit.cjs broke 31 tests in one suite that built its fixture from a hand-written dependency list, so the new sibling was absent and the copied script could not load. copyScriptWithDeps walks the require graph and exists for exactly this class - #3412 paid the same bill when one new require broke 82 tests across two suites. Migrating rather than adding another copyFileSync line keeps the class closed. The other nine suites referencing that path were triaged; none copies-and-spawns, so none needed migrating. * fix(#3906): enumerate the new shipped file, drop a vendor name from shipped data, and fix three test defects install: scripts/lib/exit-code-registry.cjs was missing from GSD_SCRIPTS_LIB_FILES, so it shipped to every install and orphaned on uninstall. The registry gave HOOK_DENY a meaning naming one harness, and that string ships into every runtime's tree - a guard correctly caught it leaking into the hermes and qwen installs. The registry is runtime-neutral infrastructure; the vendor name belongs in the ADR, not in shipped data. Two more fixture harnesses built their trees from hand-listed dependencies and broke on the new require; both migrated to the derived helper, and all 23 copy-and-spawn candidates were enumerated so the class is closed rather than patched. One generator test used a fixture code that collided with a real allocation, so the generator correctly reported a duplicate where the test expected drift. The large-payload test embedded a 256KB literal in the child's argv, exceeding Linux's 128KiB MAX_ARG_STRLEN so the child never started - it now builds the payload inside the child. * chore(#3906): backfill changeset pr number * docs(#3906): document the exit-code contract selector P2 is the first phase of this epic with a user-invocable surface, so the flag and env var owe a reference entry. Records what actually differs between v1 and v2 today (one outcome), that an unrecognized value is rejected rather than silently defaulted, and the fail-safe property that makes switching safe. --------- Co-authored-by: sim <sim@local>	2026-08-27 03:31:02 -04:00
Tom Boucher	39673ae9ff	fix(#3738): antigravity global skills/agents install to ~/.gemini/config (#3921) ... * test(#3738): antigravity global skills/agents must resolve under ~/.gemini/config Regression tests (RED first): --skills-root and gsd-tools query surfaces, install-plan dest dirs, and converter skills-path rewrite. * fix(#3738): antigravity global skills/agents install to ~/.gemini/config Antigravity's machine-local discovery scans ~/.gemini/config/{skills,agents}; the configHome (~/.gemini/antigravity) is deprecated for artifacts. Declare the ADR-1239 skills/agents 'home' override on the antigravity global layout — the same mechanism codex uses (.agents) — and divert ~/.claude/skills/ references in converted global content to ~/.gemini/config/skills/. configHome, settings, probe/migration semantics, and the local .agents layout are unchanged. * fix(#3738): retire deprecated configHome artifacts via installer migration 010 Next install converges an existing antigravity install: manifest-managed skills/gsd-*/ and agents/gsd-*.md under the configHome (a location AGY does not scan) are removed — modified files backed up first, unmanifested and non-gsd entries preserved — and now-empty containers retired. Global scope only; the local .agents surface is live. Docs + inventory updated. * fix(#3738): converter sync in bin/install.js, harness emit-root coverage, migration baseline - bin/install.js converter gains the same ~/.claude/skills → ~/.gemini/config/ rewrite as src (ADR-1508 dual copy must stay in sync). - Parity-manifest walk covers home-override emit roots (extraEmitRootsFor) so antigravity's emitted skills/agents stay differential-visible at their new install root; install-tree fixture regen confirms an unchanged key set. - skills-from-commands rule declares the antigravity converter as a runtime-scoped transform; one ack fragment covers the identity-classed workflow whose antigravity copy embeds the old skills path. - Migration 010 checksum baseline + home-override set doc updated; existing tests updated to the #3738 contract (global dest, golden parity via layout dest, integration expectations). * fix(#3738): tolerate an absent extra emit root on baseline-side measurement The base tree's installer predates the home override, so <HOME>/.gemini/config does not exist there; walk() threw ENOENT and the in-job baseline build failed. An absent extra root is the legitimate pre-override shape — skip it. * fix(#3738): review findings — manifest agents root, bare skills-path rewrite, guard comment - writeManifest resolves the agents-kind home override (_kindDestDirSafe), so the manifest records agents at their actual install root and drift detection keeps working (isolated review finding 1, major). - Converter bare forms ~/.claude/skills and $HOME/.claude/skills (no trailing slash) divert to ~/.gemini/config/skills instead of falling through to the retired configHome path (finding 2). - real-home-guard comment updated: antigravity's global agents kind is the first agents-kind home override (finding 3, doc-only). - Regression tests for both behavioral findings. * chore(#3738): changeset fragment (pr number backfilled after PR creation) * chore(#3738): backfill changeset PR number (3921) * fix(#3738): sandbox HOME in tests that install antigravity global artifacts antigravity is the first home-override runtime in the golden-parity and skills-wrapper suites (codex is not in their runtime lists), so those tests never needed HOME sandboxing — the real-home guard now (correctly) refuses their un-sandboxed global installs on CI, where HOME is the passwd home. * fix(#3738): stop the K3 sequential-sandbox env leak; sandbox L2's home-override plans K3's two back-to-back sandboxHome calls leave HOME pointing at the first sandbox once the after-hooks restore (each call saves the env as it found it, so the second saves the first's sandbox as 'original'). On the windows matrix that leaked gsd-k3-qwen-* home into the L2 property, whose antigravity/global run then (correctly) refused via the #3712 real-home guard — antigravity is the runtime that made L2's plan escape into os.homedir(). K3 now manages the env with a single restore; L2 sandboxes HOME per run, mirroring L1. * fix(#3738): L2 property's HOME sandbox must exist on disk The #3712 guard's sandbox exemption fails closed when identify(effectiveHome) is 'absent' — L2 never created its configDir, so on the windows matrix (tmpdir under the real home) the antigravity/global run refused even with HOME sandboxed. Create the per-run sandbox dir and clean it up. --------- Co-authored-by: sim <sim@local>	2026-08-27 02:24:03 -04:00
Tom Boucher	e20744eacb	enhance(#3884): failure is a value — strict argv, and --pick that signals absence (#3922) ... * test(#3884): failing-first coverage for strict argv and absence-signalling --pick ADR-3473 §8.4 says failure is a value. Three families currently encode failure as success, and this commit pins each one RED before the fix lands. Measured on this tree, 2026-08-26: gsd-tools generate-slug "test" --pick nonexistent -> empty stdout, exit 0 (#3365) gsd-tools audit-open --pick nonexistent_field -> dumps the entire human-readable audit report, exit 0 gsd-tools generate-slug "Hello World" --raw --pick bogus -> prints "hello-world", another field's value, exit 0 gsd-tools query state.planned-phase 3 (positional, no --phase) -> exit 0; STATE.md's "Phase: 2 of 5 (Widget Support)" is overwritten to "Phase: null - READY TO EXECUTE" and the frontmatter gains a corrupted current_phase_name (#3358) tests/pick-flag.test.cjs:27 previously asserted the #3365 defect as the contract ("returns empty string for missing field", success === true). That assertion is replaced by the required behavior rather than deleted. The new parseNamedArgs block calls the spec-object signature that does not exist yet, so it fails today by construction. The 11 existing behavior-lock tests are left untouched here; they are corrected in the implementation commit. C1/C4 assert at the consumer's output - STATE.md's bytes - per ADR-3180 Decision 4(b). A unit assertion on the parser would have passed throughout this defect's life. Design: .gsd/phase/feat-3884-failure-is-a-value/40-design.md Test matrix: .gsd/phase/feat-3884-failure-is-a-value/50-test-matrix.md Refs #3884 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * enhance(#3884): failure is a value — strict argv, and --pick that signals absence Implements ADR-3473 §8.4. Absence, emptiness and failure stop being interchangeable ways to say "I could not answer". parseNamedArgs (src/command-arg-projection.cts) Takes a spec object with a REQUIRED `positionals: number | 'rest'` and returns the hub's Result shape instead of a bare Record. Declaring the positional arity is what makes #3358's call site unrepresentable rather than merely detectable: an unrecognized flag or a token past the declared boundary is now InvalidArgs, naming the offending token and listing the accepted flags. The legacy positional-array call shape throws a TypeError — an internal invariant violation per ADR-3473 Decision 2, so a stale hand-written .cjs call site fails loudly instead of destructuring undefined off a Result. parseNamedArgsOrExit projects a failure onto the caller's error(); it is a projection over the one parser, not a second parser. Measured before, against a STATE.md with a populated phase-2 block: query state.planned-phase 3 (positional, no --phase) -> exit 0; "Phase: 2 of 5 (Widget Support)" overwritten to "Phase: null - READY TO EXECUTE", frontmatter gains a corrupted current_phase_name After: exit 1, `unexpected positional argument "3"`, STATE.md byte-identical. The flag form is unchanged and still updates STATE.md. --pick <field> (gsd-core/bin/gsd-tools.cjs) extractField returns {found,value}, and the pick block no longer shares one catch between "output was not JSON" and "field was absent". An absent field exits 1 with pick_field_absent, naming the field and the keys that do exist; non-JSON output exits 1 with pick_output_not_json instead of dumping the command's entire output. A field that is PRESENT with value null, '', 0 or false still prints at exit 0 — that is an answer, not a failure, and it is what keeps `--pick count` printing 0 on a fresh project. Measured before: `audit-open --pick nonexistent_field` printed the whole human-readable audit report at exit 0, and `generate-slug X --raw --pick bogus` printed "hello-world" — a different field's value, confidently, at exit 0. ADR-3409 Decision 7 explicitly deferred this contract fix to #3473; this is it. The sub-issue's "returns 0 when the count is zero OR absent" wording is superseded by the ADR rule it implements: zero prints 0, absence exits non-zero. Defaulting absence to 0 would demote "could not answer" to "the answer is zero" — the hazard docs/how-to/resolve-unreachable-guard-findings.md already warns against. Guard ledger (ADR-3473 Decision 6) scripts/lint-unreachable-guard-drift.cjs Detector A is RETIRED. Its premise — that a `--pick ... || echo` arm can never fire — is now false, so the shape it forbade is the correct idiom and keeping it would forbid the fix. Detector B (glob-consuming cat/ls, a nullglob mechanism this change does not touch) is retained in full, as are the shared scanner, the escape-marker parser and the baseline. Net: -1 detector, 0 added. The file is not deleted. Call-site audit 45 prompt-layer --pick invocations, every one a plain X=$(...) assignment — none in an if test, && chain, or a pipeline whose status is consumed, and no shell block in workflows/commands/agents/references sets -e. Of the 13 (command, field) pairs the prompt layer reads, 10 are always present; the 3 sometimes-absent ones each sit behind a prior found/existence check. No ADR-3409-class "field the command never produces" remains. Design: .gsd/phase/feat-3884-failure-is-a-value/40-design.md Test matrix: .gsd/phase/feat-3884-failure-is-a-value/50-test-matrix.md Refs #3884 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3884): escape untrusted tokens in diagnostics, and cover five unpinned rows Two review findings, both fixed here rather than recorded as limits. 1. A newline in an untrusted token forged a second stderr line. Before, plain-text mode: $ gsd-tools query state.planned-phase $'foo\nError: forged second line' Error: unexpected positional argument "foo Error: forged second line" After: Error: unexpected positional argument "foo\nError: forged second line" --json-errors mode was never affected — io.error runs that payload through JSON.stringify. Plain-text mode writes 'Error: ' + message verbatim, and the three new InvalidArgs reasons plus the two new --pick diagnostics all interpolate a token that comes straight from argv. Fixed with ONE shared helper, formatDiagnosticToken (src/io.cts), applied at every interpolation site — not a copy per site. It is deliberately NOT applied inside error() itself: several callers in this tree emit intentional multi-line diagnostics, and escaping newlines there would mangle them. The available-top-level-keys list needed the same treatment for a reason the review did not anticipate: `frontmatter get <file>` reads an ARBITRARY user document and echoes that document's own keys into the diagnostic. Verified reachable — a frontmatter key containing a newline reaches the key list — so formatKeyForDiagnosticList is guarding a live path, not a hypothetical one. Ordinary keys still render plain and unquoted; a fix that merely dropped the key would also have passed a "one line" assertion, so the test pins the escaped key's presence too. 2. Five behavior-table rows were implemented but nothing pinned them: B7 a dotted path that dies partway B9 bracket syntax on a non-array B10 a negative array index, in and out of range B14 a JSON root that is not an object B17 an @file: payload over 50KB B17 is the load-bearing one. output() writes @file:<path> instead of inline JSON past 50000 characters, and --pick resolves that BEFORE parsing; with no test, a future reordering of those two steps turns every large result into a false pick_output_not_json. The fixture seeds 1200 phase directories and measures the payload at 62474 characters, asserting the spill actually happened rather than assuming it. Refs #3884 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3884): correct the strict-argv surface against a full verification run The first full run came back with 90 failures across 12 files, none in the new tests. They were the argv surface telling me what it actually is. Ten root causes; each classified before anything was changed. I over-implemented, and that is reverted. ADR-3473 §8.4 says parseNamedArgs rejects "unrecognized and positional tokens". It says nothing about a value flag whose value is missing. Making that an error was my design decision, not the rule, and it broke a deliberately recorded contract: `--prd` with no value resolving to null (tests/init.test.cjs emptyPrdValueIsFalsyAndTreatedAsAbsent, row B5; tests/section-manifest-init-facts.test.cjs "flag-shaped value"). The "requires a value" branch is deleted outright rather than kept behind an option — an unused strictness mode is speculative generality. Unknown-flag and unexpected-positional rejection, which is what §8.4 actually mandates, is unchanged. --wave needed a third flag kind the original design did not anticipate. `--wave N` is documented (commands/gsd/execute-phase.md:4,48) and the shipped workflow reconstructs and passes it (execute-phase.md:84), while #2932 records token-PRESENCE semantics: the CLI cares only that the flag appeared, and the value belongs to the workflow layer. That is neither a boolean flag nor a value flag, so `optionalValueFlags` now exists — presence-only in `data`, and the validation cursor consumes a following non-flag token so it is not reported as a stray positional. Every other declared boolean flag was checked against every argument-hint and prose usage in commands/, workflows/, agents/ and docs/; `--wave` is the only one of this shape. Five tests were pinning forms that never worked. tests/adr857-core-without-capabilities.test.cjs passed `init plan-phase --phase 01-stub`, but the documented form is positional (docs/CLI-TOOLS.md:776) and the handler reads args[2] — which for that form is the literal string "--phase". Measured on the pre-fix build against a real .planning/phases/01-stub/ directory: init plan-phase 01-stub -> phase_found=true init plan-phase --phase 01-stub -> phase_found=false The test asserted only exit 0 and key presence, so it had been green while proving nothing about phase resolution. Corrected to the documented form and strengthened to assert phase_found === true. Same class in state.test.cjs (`--plan-count`, a flag that does not exist; the real one is `--plans`), milestone-archive.test.cjs (`init new-milestone --json`, silently ignored), and concurrency-safety.test.cjs (a bare positional field name whose OR-assertion passed because a whole-document dump happens to contain the substring it looked for). Six handlers had no argv validation at all — the same #3358 shape this phase exists to close, found while fixing the rest: init verify-work / phase-op / review / todos / remove-workspace read args[2] with nothing checking the rest, and validate health read --repair/--backfill through a bare args.includes() scan that bypassed the parser entirely. All now go through the seam, so the flag has one owner. tests/init-debug.test.cjs rows C4/C5 asserted that an unrecognized flag must NOT fail. That is the behavior §8.4 removes, and Decision 8 says a caller's local expectation does not override §8, so they are inverted and renamed — a test still called "ignores an unrecognized flag" while asserting rejection would be its own defect. Row C6's point is its PWNED canary; that assertion is kept verbatim and only its exit-status expectation changed, because the hostile token is now rejected rather than absorbed. The blast-radius estimate in 40-design.md is corrected rather than quietly left wrong. get_impact reported MEDIUM / 8 symbols upstream, and that was accurate for what the graph can see — parseNamedArgs's callers. It cannot see that those callers' handlers accept argv shapes wider than the code reading args[2] suggests, which is where the real surface was. Refs #3884 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3884): withdraw the validate-health tightening, finish the A2/A3 revert Second full run: 46 failures, down from 90. Four causes, two of them mine. Reverted `validate health` entirely — it was scope creep, and it broke a real flag. ~30 of the 46 read `unknown flag "--json"; accepted: --repair, --backfill`. The previous commit routed `validate health` through the parser on the reasoning that a flag should have one owner. That was wrong twice over: §8.4 names parseNamedArgs and count queries, and `validate health` was never a parseNamedArgs call site — it read its flags, just not through the parser, so it had no silent-drop defect to fix. Tightening it omitted `--json`, which the health-diagnostic suites use heavily. The handler is now byte-for-behaviour back to its pre-branch form. `validate context` stays converted: it genuinely was a call site, and its `--json` is now declared rather than read by a second `args.includes` scan. The five handlers that had NO validation at all — init verify-work / phase-op / review / todos / remove-workspace — stay fixed. Those read args[2] with nothing checking the rest, which is the #3358 shape this phase owns. Finished the A2/A3 revert. Three tests still encoded the deleted "a value flag with a missing value is an error" rule, including one added by the previous commit for that rule. All three now assert the reverted null contract, and the ones whose titles said "rejected" are renamed — a test named for a contract it no longer asserts is its own defect. `--wave=` and `--wave --weird` are correctly rejected. Neither is documented in commands/gsd/execute-phase.md, gsd-core/workflows/execute-phase.md or docs/, and neither is emitted by the shipped prompt layer, so both are unrecognized tokens that §8.4 mandates rejecting. `doesNotConsumeFollowingFlagAsWaveValue` keeps the property it exists for — asserted directly now, at the parser, that `--wave` does not swallow a following flag as its value — and only its exit-status expectation changed. A contradiction inside this branch, surfaced by the audit and resolved the safe way. Two pre-existing #3573 tests call `state begin-phase '2'` and `state planned-phase '2'` with a bare positional, relying on the old permissive parser to ignore it. This branch's own #3358 regression test requires that exact argv to be REJECTED. The two are mutually exclusive. Widening the router to accept a bare positional — mirroring complete-phase — would have silently re-opened #3358, and was verified to do exactly that: with the widened router, `query state.planned-phase 3` returned exit 0 and wrote current_phase_name again. It is reverted. docs/CLI-TOOLS.md:116 and docs/COMMANDS.md:2192 document only the `--phase N` form for both verbs, so the two #3573 tests move to it. Their assertions were never about the call shape — only that total_phases survives the resync — and both still pass. complete-phase is untouched: its bare positional IS documented, and it keeps the dynamic boundary and the negative-space note that record why. The audit that produced this is in the PR body: for every handler whose declaration changed, the flags it reads anywhere in its body, the flags the shipped surface documents, and the shapes the suite passes, compared. The `--json` miss was a pattern, not an accident — declaring a handler's flags from its parseNamedArgs call alone misses whatever it reads elsewhere. Refs #3884 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#3884): backfill the changeset PR number Refs #3884 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-27 00:12:13 -04:00
Tom Boucher	878f25025c	enhance(#3905): the exit-code registry — one number, one meaning, enforced at build (#3920) ... * feat(#3905): the exit-code registry — one number, one meaning, enforced at build A generated registry replaces locally-invented exit codes. Every entry records code, name, meaning, owning module and the decision that authorized it. The generator refuses to build a table where two entries claim one code, two claim one name, a code falls in a range Node or the shell reserves, 2 is claimed by anything but the hook adapter, or an allocation carries no justification. exitCodeFor is pure and total: it throws rather than returning undefined, including for prototype-chain names. Inert by design — nothing emits a registered code until #3906. Every registered code is non-zero, asserted over the whole table, so a caller testing for failure behaves identically for pass and trips for everything else. * feat(#3905): make the registry generator's failures machine-readable Adds a --json mode carrying {ok, reason, context, detail}, where context is a typed payload naming the specifics the prose embedded - which code collided and under which names, which band rejected a code, which field was missing. The tests now assert on that structure instead of regex-matching the generator's stderr, which CONTRIBUTING prohibits, and the CONTEXT.md glossary gains the entry the issue's scope requires. * test(#3905): refresh the install-tree fixtures for the new declaration The registry declaration ships in the install tree, so all 19 golden fixtures needed regenerating. Caught by the remote matrix, not by lint:ci - the install-tree goldens are verified by a test rather than a lint, so a newly shipped file clears every local gate and fails only under the suite. * chore(#3905): backfill changeset pr number --------- Co-authored-by: sim <sim@local>	2026-08-27 00:10:11 -04:00
Tom Boucher	7e9d33c378	enhance(#3915): stryker on the official tap runner, 29m to 12m on the critical path (#3919) ... * enhance(#3915): stryker on the official tap runner, per-test coverage The 'command' runner is the one runner Stryker excludes from coverage analysis, which forced coverageAnalysis:'off' and made mutation cost strictly linear in (mutants x whole-shard test time). The frontmatter shard measured 1751s against 212s for the next slowest. Swap to @stryker-mutator/tap-runner (official plugin, peer-pinned to the @stryker-mutator/core@9.6.1 already installed) and turn coverage analysis on, so Stryker re-runs only the test files that cover each mutated line. Per-shard injection moves from a MUTATION_TEST_CMD command string to MUTATION_TEST_FILES, read by a new fail-closed resolveMutationTestFiles() that mirrors resolveMutationBreak. It stays derived from scripts/mutation-matrix.cjs and now has a single owner: the union moves behind allCoveredTests() and stryker.config.mjs no longer imports COVERED. The resolver existence-checks every entry, because the tap runner resolves tap.testFiles with glob() and a non-matching pattern yields an empty list silently - a fast, confident, meaningless run. tap.forceBail is off by measurement, not preference: a structural AST audit found 3 of 26 shard test files spawn subprocesses, and bail fires on every killed mutant, so leaving it on would kill processes mid-spawnSync and orphan their children. The matrix 'isolation' field is removed - per-file process isolation is inherent to the tap runner, so the field had no consumer left. Score arithmetic is unchanged and now pinned: mutationScore counts NoCoverage in the same denominator as Survived, which both Stryker's thresholds.break and check-mutation-score-ratchet.cjs read. A new non-vacuity test proves it diverges from mutationScoreBasedOnCoveredCode, so the gate cannot be quietly swapped to the field that would make every floor trivially satisfiable. Refs #3915 * fix(#3915): enforce the resolver's documented containment contract Review findings, all fixed in place. resolveMutationTestFiles claimed to verify each entry exists 'relative to the repo root' but used a bare fs.existsSync(path.join(...)), which accepts an existing DIRECTORY and lets ../ segments escape the root (path.join('/repo/root','../../etc/passwd') resolves to /etc/passwd). Not reachable from PR content - the value only ever comes from the static COVERED registry via CI env - but a fail-closed contract that overstates its own guarantee is a defect in the contract. Each entry is now resolved, rejected if path.relative puts it outside the root, and required to be a regular file. Three hostile-input tests added; the original missing-file wording is preserved so the existing assertion still binds. Also: removed a stale buildResult comment still naming the isolation field this branch deleted; hoisted one top-level path require in place of three inline ones; de-duplicated the derived-test-list expression in the tests to a single const, deliberately still re-derived from COVERED rather than calling allCoveredTests() so the assertion cannot become a tautology; tightened the workflow-parity assertion to exact equality; changed the tap-runner range from an exact 9.6.1 to ^9.6.1 so it tracks the caret-ranged core its peerDependency pins exactly. Regenerated examples/dynamic-context-management/CONTEXT-INDEX.json, which the earlier CONTEXT.md edit left stale - lint:ci was red on lint-example-parser-parity until it was refreshed. Refs #3915 * test(#3915): kill model-catalog survivors, set frontmatter budget from measurement From mutation run 33026833181 (all 13 shards, dispatched on this branch before any PR). WALL TIME: frontmatter measured 713s (11m53s) vs 1751s (29m11s) on the command runner, a 59% cut. timeoutMinutes 60 -> 20 (1.68x measured). Not deleted outright: the shared 15-minute default would leave only 21% headroom, and this module's mutant count grew 1.8x in one change. MODEL-CATALOG: came back 57.91 against its floor of 58. Diagnosed from the report JSON, not assumed - all 24 of its new RuntimeError mutants are in the load-time catalog bootstrap, so mutating them makes the module throw at require. Under node --test that is a failed test file (Killed); under the tap runner the process dies before emitting TAP, which Stryker classifies RuntimeError and excludes from the denominator. Add them back as killed and the shard is 248/416 = 59.62, the pre-change number exactly. Detection did not regress, classification changed. The floor is NOT lowered to absorb that. 11 new behavioural tests target ~42 genuinely surviving mutants: exact Set equality on the EFFORT_RENDERING/EFFORT_ARGV supported sets, an exact-string table render that makes the column-width arithmetic observable (padEnd never truncates, so the old substring assertion could not see a too-small width), clampEffortForHost's full null matrix plus a spoofed-toString host, and a prototype-pollution guard driven through Object.fromEntries. Mutants judged equivalent were skipped rather than papered over; reasoning is in the phase artifacts. All 15 new assertions verified against unmutated code. lint:ci exit 0. Refs #3915 * test(#3915): ratchet model-catalog's floor to 74 on measured 75.26% CI run 33029755081 measured model-catalog at 75.26% (295 killed / 49 survived / 48 no-coverage / 24 runtime-error, totalValid 392), so check-mutation-score-ratchet.cjs correctly failed the shard for unclaimed headroom: 17.26 points above the declared floor of 58, well past the 5-point slack. Floor raised 58 -> 74 (floor(75.26)-1, this file's documented convention), with RATCHET_BASELINE updated in the same diff as the equality assertion requires. Worth recording why the number moved this far. The shard first came back at 57.91 under the tap runner and the temptation was to lower the floor to match. The drop was not a regression: all 24 of its RuntimeError mutants sit in the load-time catalog bootstrap, and adding them back as killed reproduces 248/416 = 59.62, the pre-swap figure exactly. Rather than absorb a reporting artifact by weakening the gate, 11 behavioural tests went after the genuinely surviving mutants and killed 68 of them - carrying the module from 59.62 past its old ceiling to 75.26, within reach of the ADR-456 target of 80. The #3007 measurement is kept as clearly-labelled prior context rather than deleted, so the entry does not read as carrying two current numbers. Refs #3915 --------- Co-authored-by: sim <sim@local>	2026-08-26 22:43:14 -04:00
Tom Boucher	8edace40d5	enhance(#3904): one exit module — generate the scripts-side copy from a single source (#3917) ... * test(#3904): failing-first coverage for the drifted scripts-side exit module The scripts/ copy of the CLI exit seam has no json-error arm, so an unexpected throw prints a raw stack where the documented contract promises {ok:false,reason,message}. Adds the consumer-altitude reproduction plus the negative space it must not swallow, the one-cell assertions for json-error mode, and the standalone-load constraint. RED until the generator lands. * enhance(#3904): generate the scripts-side exit module from one source src/cli-exit.cts becomes the single source of truth and scripts/lib/cli-exit.cjs a generated artifact of its compiled output, byte-compared by a --check entry in lint:generated-sync. The two had drifted: only the .cts copy emitted the documented {ok:false,reason,message} envelope on an unexpected throw, so a scripts-side tool printed a raw stack where docs/json-errors.md promises structured output. The generated file is committed and must load on an unbuilt clone (64+ consumers, incl. check-env.cjs), and gsd-core/bin/lib/cli-exit.cjs is gitignored tsc output that doubles as the build sentinel, so it cannot be required from there. The exit module therefore drops its io.cjs import: the json-error-mode accessors move into it and io.cts re-exports them, leaving its export surface unchanged. The flag lives in a Symbol-keyed cell on globalThis because one source emitted to two locations means two module instances, and a module-level flag would give them two independent values. * chore(#3904): changeset for the generated scripts-side exit module * docs(#3904): name which surfaces honor the json-error envelope contract docs/json-errors.md described the structured envelope as what runMain does without saying which copies of runMain actually had the branch — a claim that was silently false for every scripts/-side tool. Also drops a redundant source-grep test whose marker grew the unverified allow-test-rule pool past its ceiling; the behavioral test beside it proves the same property through real module resolution. * test(#3904): compare exit verdicts, not stderr bytes, across the two copies The parity test asserted byte-identical stderr, which the plain-text path cannot satisfy: the generated copy carries an 11-line banner, so its stack frames report line numbers offset by exactly that much, and the path normalizer stopped at the colon. Byte-identical stack traces were never the contract - two files at two paths necessarily differ there. Now compares the parsed envelope under json mode, the first line and exit code on the stack path, and exact output for ExitError. * chore(#3904): backfill changeset pr number --------- Co-authored-by: sim <sim@local>	2026-08-26 21:24:34 -04:00
Tom Boucher	6b7df61938	enhance(#3881): one YAML parser — vendored js-yaml replaces the hand-rolled dialect (#3888) ... * docs(#3881): answer §8.1's open question and correct three wrong premises ADR-3473 §8.1 carries a blocking open question with a forcing function: it must be answered before any implementation PR for the rule opens. Answered here as (a), a string-coercing adapter, with the measurement that settles it. The sequencing note bet that §8.8's schema would make (b) tractable. Measured against merged reality it does not: only 33 of extractFrontmatter's 78 non-test call sites read STATE.md, and two of the five compensating mechanisms §8.1 lists survive real types, leaving ~31 lines across 3 call sites as the actual prize. Also corrects three claims verified false while answering it. §8.1's justifying sentence names #3349 and #3360 as defects a real parser would fix; both are already fixed on next, confirmed by executing the compiled parser rather than reading it. The guard roster calls lint-frontmatter-scalar-broad-grep.cjs an expected casualty of this rule, but it guards shell grep idioms in workflow bash fences and never touches our parser. The same roster calls lint-vendored-deps.cjs reusable as-is; it is hardcoded to re2js throughout. The last two were caught by applying the rule this amendment records -- a factual claim in this ADR is a hypothesis until the implementing phase executes it -- on its first use. Refs #3881 * docs(#3881): record that §8.1's fork is ill-posed and (a) is not implementable An adversarial pass on the Phase 4 design established by execution that extractFrontmatter is not a YAML parser but a line-oriented scanner whose output is a function of raw source text. Four spellings of the same value collapse to one js-yaml tree but produce four distinct legacy strings, one of them mangled. No adapter over a tree can choose among outputs the tree does not distinguish, so fork (a) -- keep a string-coercing adapter so the existing contract holds -- cannot be built. For any document with a non-scalar value, (a) collapses into (b); about 26 percent of frontmatter-carrying documents have one. Also records three design defects and one new attack surface, all confirmed by execution: catching a parse failure and returning {} would delete the frontmatter block on the next write at eight call sites that conflate empty with unparseable; an empty value yields null where legacy yields {}, and reconstructFrontmatter omits null-valued keys, so the shipped state template's empty progress key would vanish; the #1882 truncation probe is parseYamlRegion itself rather than a pre-parse heuristic, so it cannot both stay unchanged and survive that deletion; and FAILSAFE_SCHEMA still resolves aliases, expanding seven lines to 22.8 MB. The rule is not deferred. The measurement is the deliverable and the re-scoping is recorded as an open question with a forcing function, per section 8's own rule. Refs #3881 * test(#3881): failing-first rows for block scalars, unicode keys and the missing #3594 matrix Creates tests/feat-3594-parser-adversarial-frontmatter.test.cjs, the file the fixture README instructs contributors to register fixtures in but which never existed. Section C: table-driven ownership check over tests/fixtures/adversarial/frontmatter/ so a fixture with no matrix entry fails loudly; six existing fixtures (duplicate-keys, crlf-mixed, unclosed-block, unicode-keys-and-values, null-byte-value, huge-bounded) each get the invariant its README states. B1 blockScalarValueIsNotTheBlockIndicator: parsing commands/gsd/add-tests.md must give argument-instructions the instruction text, not the literal '|'. RED today. B2 blockScalarDoesNotInventATopLevelKey: same parse must not produce a top-level Example key scraped from inside the block body. RED today. B3 unicodeKeyRoundTripsAsIs: the 相 key in unicode-keys-and-values.md must survive parsing; today it is silently dropped. RED today. Refs #3881 * chore(#3881): vendor js-yaml and generalize the vendored-deps guard to a manifest Packaging step for ADR-3473 §8.1: makes js-yaml available to gsd-core/bin/** without promoting it out of devDependencies (promoting broke every installed tree, #3496). gsd-core/bin/lib/vendor/js-yaml.cjs is a verbatim copy of node_modules/js-yaml/dist/js-yaml.js (the self-contained UMD dist bundle, not index.js), exposing load/dump/FAILSAFE_SCHEMA/YAMLException with zero require() calls of its own. src/vendor/js-yaml.d.cts is hand-authored, not copied, because js-yaml ships no upstream .d.ts and @types/js-yaml is not installed. It is deliberately narrow, declaring only the four symbols in use, so anchors/aliases/custom types/loadAll are unreachable from typed code -- a compile-time enforcement of ADR-3473 §8.1's refusal to expand alias resolution for security reasons. Because it has no upstream counterpart it is excluded from the byte-compare. scripts/lint-vendored-deps.cjs is refactored from a script hardcoded to re2js into a table-driven VENDORED manifest (one row per package: upstream/vendored .cjs paths, optional .d.cts paths, twin kind upstream-verbatim vs hand-authored) so a second vendored package does not require a second hardcoded check block, per ADR-3473 §8.3 'one implementation per rule'. The four existing re2js checks (vendored .cjs vs node_modules, vendored .d.cts vs node_modules, src/vendor twin vs bin-side twin, devDependency version pin vs installed version) are preserved unchanged; verified pass/fail identical before and after the refactor, and the guard's ability to fail was re-proven with a deliberate one-byte append to both re2js.cjs and js-yaml.cjs, then restored. docs/INVENTORY.md and docs/INVENTORY-MANIFEST.json (via gen-inventory-manifest.cjs --write, run after build:lib) register vendor/js-yaml.cjs. gsd-core/bin/lib/vendor/README.md documents both vendored packages and the two twin kinds. Refs #3881 * feat(#3881): parse .planning frontmatter with the vendored js-yaml ADR-3473 §8.1: extractFrontmatter's read path is no longer a hand-rolled line scanner. parseYamlRegion, escapeDoubleQuoted, unescapeDoubleQuoted and parseQuotedScalar are deleted (not patched); parsing now goes through the vendored js-yaml (./vendor/js-yaml.cjs) under { schema: FAILSAFE_SCHEMA, json: true }. Everything js-yaml does not do is layered on top, in one place, carrying the seven design-doc consequences: 1. Empty value: a null js-yaml value is coerced to {} (matching legacy's own empty-value contract) so reconstructFrontmatter — which omits null-valued keys — still round-trips a bare `key:` line instead of deleting it. Verified live: progress: with no value survives parse -> reconstruct -> re-parse. 2. Unparseable no longer collapses to a bare {}: a new FRONTMATTER_UNPARSEABLE Symbol (exported), keyed exactly like the existing #3257 FULL_LINE_COMMENTS channel, is carried on the {} returned for malformed/refused YAML. Invisible to Object.keys/entries/JSON.stringify/for-in, so the 70 call sites that never inspect it are unaffected; wiring the 8 hasFrontmatter sites to consult it is a separate change, not done here. 3. Non-scalar object-list items (the four spellings of `- test: a b` that js-yaml collapses into one tree shape) are rendered as a canonical `key: value[, key2: value2]` string per item, keeping the existing array-of-strings value SHAPE. A full corpus differential over all 1702 tracked markdown files found 11 residual divergences from the legacy parser (enumerated in the PR/report), most of them the parser now being MORE correct (a dropped quoted top-level key, the block-scalar/phantom-key defect, a dropped Unicode key). 4. The #1882 truncation probe still runs the one real parser, but derives its key count from js-yaml's own thrown error and mark.line when the whole region doesn't parse cleanly (the dominant real truncation shape: fence opened, well-formed keys, no closing fence). Verified against both the clean-parse and the exception-fallback path. 5. The #3257 comment channel now attributes each pending column-0 comment against js-yaml's own parsed top-level key list (matched by literal key text, in document order) instead of the legacy ASCII-only key regex, so a comment above a Unicode key attaches correctly. 6. Anchors, aliases and merge keys are refused outright (a raw-text pre-scan, since FAILSAFE_SCHEMA still resolves them) — corpus occurrences today: zero. A 7-line billion-laughs fixture is verified refused rather than expanded. 7. A literal U+0000 is swapped for a private-use sentinel before the parse and restored in every resulting string afterward, since js-yaml rejects NUL unconditionally under every schema. escapeDoubleQuoted is deleted and reimplemented via js-yaml's dump() (forced double-quoted style), with control-char hex escapes lowercased to keep serialized output byte-stable (#1779 emitted lowercase); it keeps its exported name and signature for its two other call sites (commands.cts, runtime-artifact-conversion.cts), which need no change. frontmatterDeepEqual, the comment channel, sliceTopLevelFrontmatterSegments, regenerateFrontmatterKey's guard, noOpObjectListSetError and parseMustHavesBlock are all unchanged — retiring them is fork (b) and is not this phase. Refs #3881 * fix(#3881): quote template placeholders and preserve unparseable frontmatter SECURITY.md/UI-SPEC.md/VALIDATION.md wrote frontmatter placeholders as bare {N}/{phase-slug}/{date}, which is valid YAML flow-mapping syntax under the vendored js-yaml parser, not the literal placeholder text intended. Quote them so they parse as strings. Wire the FRONTMATTER_UNPARSEABLE Symbol (exported but unused) at the 8 call sites in state.cts/state-transition.cts that compute hasFrontmatter via Object.keys(extractFrontmatter(...)).length > 0 and reassemble the document without a frontmatter block when false. That check conflated 'no frontmatter' with 'unparseable frontmatter' (both parse to {}), so a document with a merge-conflict marker or refused alias in its frontmatter had that block silently dropped on write. Each site now preserves the exact raw bytes stripFrontmatter removed when the marker is set, leaving the genuinely-empty case unchanged. Refs #3881 * test(#3881): consequence and boundary coverage for the js-yaml migration Rows: A1 emptyValuedKeySurvivesAWrite, A2 unparseableDocumentKeepsItsFrontmatterBlock, A3 unparseableIsDistinguishableFromEmpty, A4 nonScalarValuesCanonicalize, A5 truncationProbeStillFiresOnAnOpenFence, A6 commentsStayOnTheirOwnKey, A7 anchorsAndAliasesAreRefused, A8 aliasExpansionCannotExhaustMemory, F1 UNTERMINATED_KEY_THRESHOLD boundary, F2 alias/nesting refusal bound, F3 frontmatter size boundary (huge-bounded.md + larger). Adds tests/fixtures/adversarial/frontmatter/anchor-alias-bomb.md and its entry in the feat-3594 fixture matrix. Refs #3881 * docs(#3881): document the vendored parser, correct a stale rationale, add a vendoring how-to Refs #3881 * docs(#3881): correct the frontmatter glossary entry Two errors in the entry as first written: it named parseYamlRegion as part of the read path when that function is deleted, and it recorded the eight hasFrontmatter call sites as unwired follow-on work when they were wired in e35ac2a2c. Also records the scope caveat that the CLI write path rebuilds the frontmatter block independently, so the marker binds at the transform layer. Refs #3881 * docs(#3881): record the semantic-migration decision and the counted guard ledger The maintainer chose the full semantic migration over splitting the rule into its own epic or patching the scanner, so section 8.1 is answered as "the fork was ill-posed and the migration is semantic" rather than as (a) or (b). Also replaces the pre-implementation guess that this phase would shrink the guard surface with the counted result: excluding vendored third-party lines the hand-maintained surface is net +307, and frontmatter.cts grew by 68 lines despite four functions being deleted, because the compatibility layer over js-yaml is larger than the scanner it replaced. Section 8.1's stated benefit is therefore not delivered as written; what improved is the kind of code maintained, not the amount. Decision 6 requires recording that rather than netting it away. Refs #3881 * chore(#3881): changeset for the vendored YAML parser migration Refs #3881 * test(#3881): golden parity, round-trip property and packaging coverage Refs #3881 * fix(#3881): refuse anchors structurally and fold in review findings ADR-3473 §8.1 review findings, addressed inline: Finding 1 (BLOCKER): refuseAnchorsAndAliases was a raw-line regex that matched only the bare-key spelling (key: &x). A quoted key ("a": &x), a flow mapping ({b: &x}) and a flow sequence ([&x, *x]) all define/use the SAME anchor mechanics while never matching that line shape, so the exact expansion the guard exists to stop went straight through unrefused (a 303-byte quoted-key bomb expanded to ~35.8MB). Replaced with js-yaml's own `load` `listener` callback, which reports `state.anchor` for every event belonging to an anchored node in every spelling, and throws from inside the callback to abort before any expansion (~1-2ms vs full expand-then-discard). A merge key with an alias is still refused (merge always requires a previously anchored node, so the alias itself trips the listener); a bare merge key with NO alias is no longer separately refused, documented as intentional: FAILSAFE_SCHEMA never resolves `!!merge`, so it carries no expansion risk. Table-driven tests added for all four bypass spellings + merge key, plus a quoted-key-spelled billion-laughs fixture registered in the adversarial matrix and README. Finding 2: src/vendor/js-yaml.d.cts's docblock falsely claimed anchors/ aliases were "simply UNREACHABLE from typed code" through the twin. Corrected to state the truth: anchor/alias resolution is document-level `load` mechanics reachable through exactly the declared surface, and refusal is enforced at RUNTIME (Finding 1's listener), not by the type surface. Finding 3 (MAJOR): the null-byte sentinel (U+E000) round-trip was non-injective — restoreNullBytesDeep rewrote every U+E000 in the parsed tree back to NUL, including one the document author legitimately wrote, silently corrupting it. Now refuses outright whenever the raw region already contains U+E000 (consistent with the existing anchor/merge-key refusal path), making the substitution provably injective. Tests added for a real NUL alone (preserved), a pre-existing U+E000 alone (refused, not corrupted), and both together (refused, not merged into one byte). Finding 4 (MAJOR): scripts/lint-vendored-deps.cjs's `srcTwin` field was dead for a hand-authored row (only read inside the upstream-verbatim branch) — exactly how Finding 2's stale docblock drifted unnoticed. Added checkHandAuthoredTwin: every value-level export the twin DECLARES must be an actual own property of the vendored runtime module at require-time. Tests added, including a sensor that a declared-but-nonexistent export IS caught. Finding 5: the existingFm/hasFrontmatter/stripFrontmatter/fmPrefix/ unparseableFm/reassemble preamble, copy-pasted at 7 sites in state-transition.cts plus a sixth hand-inlined copy in state.cts's cmdStateCompletePhase, is now one exported helper (beginFrontmatterReassembly) every site routes through, including the hand-inlined one. Three call sites (beginPhaseCore, patchCore, updateCore) keep a literal `body = stripFrontmatter(content)` assignment alongside the helper call so scripts/lint-state-write-path-drift.cjs's single-hop backward scan (which does not chase aliases) still sees the strip; stripFrontmatter is pure/idempotent so the extra call changes nothing observable. Finding 6: corrected the frontmatter.cts docblock's stale "wiring is a separate change" claim (the 8 call sites are wired on this branch) and the changeset's backlink from (#3473) to (#3881). Finding 7: fixed the lint:ci failures blocking the gate — an @typescript-eslint/only-throw-error violation from throwing a bare Symbol as the anchor-detected signal (now a real Error subclass), unused-var warnings left over from the Finding 5 refactor, a lint-test-file-count cap exceeded by two migration-specific test files (allowlisted with justification), and the lint-state-write-path-drift false positive from Finding 5's helper (fixed above). tests/frontmatter-golden-parity.test.cjs:117's execFileSync already carried an explicit timeout; no change was needed there. Golden fixture: added a golden entry for the new anchor-alias-bomb-quoted.md fixture ({} — matches what the legacy line scanner would also produce, since it independently dropped every quoted top-level key). No other corpus document diverges: real .planning/ documents carry zero anchors/aliases/merge keys/U+E000 today. Refs #3881 * fix(#3881): fold in second-round review findings Finding 1 (BLOCKER): tests/frontmatter.test.cjs pinned the pre-migration ASCII-only key regex for the Unicode fixture; updated to require the 相 key's value now that js-yaml has no such restriction. Audited the rest of the file for other pre-migration pins (block scalars, quoted keys, flattened values, empty values, duplicate keys, unclosed blocks, null bytes) by execution against real fixtures; found none regressed. Finding 2: parseYamlRegion and escapeDoubleQuoted renamed to parseGuardedYamlRegion and escapeDoubleQuotedScalar in src/frontmatter.cts so no function still answers to the deleted hand-rolled scanner's name (ADR-3473 §8.1 "deleted, not patched"). escapeDoubleQuotedScalar's three external call sites (src/commands.cts, src/runtime-artifact-conversion.cts) updated in the same change — a mechanical rename, not an ADR-amendment matter. Finding 3 (BLOCKER): fixed a real crash and a silent data-loss bug found by execution. A top-level key named constructor/__proto__/toString/ valueOf/hasOwnProperty crashed reconstructFrontmatter (bracket read resolving an inherited Object.prototype member); a key literally named __proto__ was silently DROPPED entirely (bracket assignment on an ordinary {} invoked the inherited __proto__ setter instead of creating a data property). Fixed by building every parsed Frontmatter object with Object.create(null), and replacing an `in` check with hasOwnProperty.call in propagateCommentChannel. Added round-trip tests for all five hostile keys, each with its own leading comment. Finding 4 (MAJOR): escapeDoubleQuotedScalar's docstring falsely claimed full byte-stability across the migration. Verified by execution: BEL/NUL/ NEL/NBSP/LS/PS/BOM now emit YAML-named escapes instead of the old hex/raw- literal forms. Proved round-trip equivalence (each escape re-parses to the exact source codepoint) and corrected the docstring. Found and fixed a related real defect while verifying: a lone UTF-16 surrogate was emitted BARE (scalarNeedsDoubleQuoting didn't trigger), producing genuinely unparseable YAML that silently collapsed to {} on re-read — extended scalarNeedsDoubleQuoting to route surrogates through the quoted+escaped path. Finding 5 (MAJOR): countKeysBeforeTruncation went silent on 4 real truncation shapes (unquoted colon, open flow collection, mis-indented sibling key, refused anchor). Root cause: the mark-based prefix recovery excluded the very line whose key needed counting, and a mark-less refusal never entered the recovery branch at all. Fixed by taking the max of two lower bounds: the longest parser-verified line-prefix, and a raw-text count of key-shaped lines (reusing the same key-shape pattern this file already uses for isFrontmatterShaped). Extended test-matrix row A5 table-driven over all 4 regressed shapes. Finding 6: the design doc's claim that no test owned the #3594 adversarial fixture corpus was false — consolidation epic #1969 had already folded it into tests/frontmatter.test.cjs. An earlier commit on this branch re-created a standalone duplicate under that false premise; folded its genuinely-new coverage (fixture-ownership check, anchor-bomb fixtures, block-scalar B1/B2 rows) into frontmatter.test.cjs and deleted the duplicate file. Corrected the false claims in 40-design.md §3.3.1 and the ADR's §8.1 note, including the roadmap-sibling claim (no such file exists). Finding 7: the golden serializer sorted object keys, making it structurally blind to the key-order-parity invariant ADR-3473 §8.1 actually claims. Made it order-preserving and regenerated the golden fixture from a standalone compile of the legacy (pre-#3881) parser at ddde001af; the current parser matches it with zero undocumented divergences, confirming key-order parity genuinely holds. Extended row A2 table-driven across 6 of the remaining 7 transitionCore kinds (all pass) plus documented, by execution, a newly-discovered 8th-site regression: state.cts's cmdStateCompletePhase calls the same preservation helper but its result is clobbered by a later unconditional resync — filed as a distinct finding rather than fixed here (touches syncAndPreserveStateMd, outside this change's verified scope). Refs #3881 * fix(#3881): preserve unparseable frontmatter through the CLI write path Characterization (executed, before/after shown): case (b), not (a). The frontmatter FENCE survives — `state complete-phase` on a conflict-marked STATE.md returns success and a well-formed, freshly-derived frontmatter block, not a document with no frontmatter at all. But the block's actual content (the merge-conflict markers, and with them any signal to a human that the document was in conflict) is silently discarded and replaced. Root cause was two clobber sites, not one: 1. syncStateFrontmatter (src/state.cts) re-parses the already-preserved `transformedContent` from readModifyWriteStateMd, finds {} + the FRONTMATTER_UNPARSEABLE marker, and unconditionally rebuilt a fresh frontmatter block from the body anyway. 2. Even after (1) is fixed, applyPostSyncPreservation's own postFm/applyStatePreservation/authoritativeFm-reassertion machinery re-extracts frontmatter from syncedContent, restores curated fields from the pre-write snapshot, and reconstructs a NEW block again — confirmed live via `state begin-phase`, which still lost the markers after fixing (1) alone. Both are now guarded by the same predicate (isUnparseableFrontmatter, checking FRONTMATTER_UNPARSEABLE): when the ORIGINAL frontmatter did not parse and the caller is not on ADR-3408 §8.3's closed "body wins" list, both functions return their input content unchanged rather than re-deriving over it. The closed list (cmdStateSync #905, /gsd-health --repair's REGENERATE_STATE, both routed only through writeStateMd, which never reaches applyPostSyncPreservation and passes sanctionedPermanentEmptyFallback=true to syncStateFrontmatter) is untouched — neither widened nor narrowed; verified by execution that `state sync` still overwrites the conflict-marked block exactly as before. Other verbs sharing the same readModifyWriteStateMd path were checked and were equally affected before this fix: state update, query state.patch, and state begin-phase all lost the conflict markers (RED, shown by execution), and all three now preserve them (GREEN). Covered table-driven in tests/feat-3881-yaml-parser-consequences.test.cjs's new A2b describe block, which drives the real CLI verbs via runGsdTools — not just the pure transitionCore layer the earlier A2 rows exercised — plus a control asserting state sync's body-wins contract is unchanged. Refs #3881 * fix(#3881): restore the parse surface's prototype and fix remote-runner failures Root cause of the bulk of the 88 remote-runner failures: extractFrontmatter/parseGuardedYamlRegion handed back Object.create(null) trees for prototype-pollution safety, but assert.deepStrictEqual compares prototypes, so every assertion against a plain object literal failed (57 frontmatter.unit.test.cjs + 5 frontmatter.test.cjs + others). Fixed by keeping the internal construction null-prototype (unchanged) and converting to a plain-prototype tree via Object.defineProperty (never bracket assignment, so __proto__/constructor/toString keys stay safe) at the parseGuardedYamlRegion/unparseableResult return boundary only; the internal FULL_LINE_COMMENTS Symbol channel is copied by reference, not recursed, so its own __proto__-safety is untouched. Per-class fixes: (1) bomAcrossArtifactTypes was the same prototype bug, no separate code change needed. (2) frontmatter-cli #1660: added objectListFieldWouldLoseData, a broader lossy-field detector alongside the existing byte-identical noOpObjectListSetError -- js-yaml's flattenObjectListItem now correctly includes every sub-key of an object-list item (a real bug fix over the legacy scanner, which silently dropped every field but the first), so a set that drops that now-included data is no longer byte-identical to the original and needs its own guard. (3) uat.test.cjs: updated the pinned expectation for the human_verification quote-stripping artifact -- js-yaml resolves quoting correctly where the legacy regex left an unbalanced quote; documented as an intentional, non-lossy behavior change. (4) smart-entry: added a fallback-only loadWithAmbiguousColonRepair so a column-0 key: value line whose value itself contains an unquoted colon (the #2571 hand-edited-STATE.md shape) round-trips instead of failing the whole frontmatter block closed. (5) frontmatter.unit.test.cjs bracket-array leniency: added a second fallback, repairMalformedInlineArrays, restoring the legacy scanner's tolerant inline-array handling (consecutive/blank commas, unclosed bracket) -- both repairs run ONLY after the primary parse already threw, so well-formed documents are unaffected. (6) prompt-injection-scan: src/frontmatter.cts had a literal U+FEFF BOM embedded in a comment illustrating the #2977 fix; replaced with the U+FEFF text escape. (7) eslint-glob-coverage: allowlisted the new src/vendor/js-yaml.d.cts vendored type declaration, same precedent as the existing re2js.d.cts entry. (8) frontmatter-golden-parity: git ls-files *.md now runs with -c safe.directory=* (process-scoped) so it survives the remote runner's dubious-ownership check without a persistent git config write. Refs #3881 * chore(#3881): backfill changeset PR number Refs #3881 * test(#3881): make golden parity resistant to unrelated tree churn A corpus-wide snapshot keyed to every tracked *.md file was coupled to mutable-by-design files: .changeset/*.md's pr:0 -> real-PR-number backfill is a required workflow step, not a parser change, yet it turned this suite red. Training people to 'just regenerate the golden' on that kind of failure defeats the point of the snapshot. Exclude .changeset/** from the golden corpus entirely, tolerate tracked *.md files with no golden entry (they postdate the capture) instead of failing on them, keep hard failures for a golden entry whose file has vanished from the tree and for any real parity divergence, and add a coverage floor so the enumeration cannot quietly degrade to comparing a handful of files. Golden regenerated by recompiling the legacy pre-migration parser (git show ddde001af:src/frontmatter.cts) standalone, independent of the current parser, over the same non-changeset corpus. Refs #3881 * test(#3881): make the parser golden hermetic instead of tree-keyed This repo merges ~21 commits/day; a 14-day sample measured 937 touches of the exact files (commands/gsd/*.md, gsd-core/workflows/*.md, agents/*.md, docs/*.md) the prior golden pinned by tracked path. Any PR editing one of those files' frontmatter for reasons unrelated to the parser (an argument-hint addition, an allowed-tools tweak) turned the suite red, and the reflex fix -- "regenerate the golden" -- overwrote the very snapshot meant to catch a real regression. Excluding .changeset/** was not enough; the design itself was wrong: a regression fixture must not be keyed to mutable repo paths, and a single 376-entry JSON every such PR touches is also a guaranteed merge-conflict surface. Rebuilt the fixture to carry its own documents: each of 51 entries stores a stable id, literal documentText (shrunk from a real ddde001af-era corpus document), and an expectedParse captured independently from the pre-migration legacy parser (git show ddde001af:src/frontmatter.cts, compiled standalone against its byte-identical sibling modules). The test reads no tracked path, shells out to no git command, and enumerates no tree -- a PR editing commands/gsd/help.md cannot affect it. Every entry's reconstruction was verified at capture time to reproduce both the current and legacy parser's output on the original document; 0 of 51 candidates were dropped by that check (1, the deliberately-unterminated unclosed-block.md adversarial fixture, has no closing fence to truncate at and is stored unshrunk). Kept the 5 documented DIVERGENCES rows (now diverges:true entries) and the D2 order-preserving structural serializer that keeps the comparison from passing vacuously; dropped the tree-enumeration helpers, the coverage floor, the post-capture-skip logic, and the vanished-file check -- all artifacts of the path-keyed design. Refs #3881 * fix(#3881): resolve vendored-deps paths independently of cwd shape Five rows in tests/lint-vendored-deps-manifest.test.cjs failed on windows-latest CI: the test passed absolute scratch-file paths into compareFiles()/checkRow(), whose helpers joined every input onto ROOT via path.join(ROOT, rel), producing garbage when the input was already absolute. It surfaced on windows-latest specifically because GitHub's Windows runners checkout the repo on a different drive than TEMP, so path.relative(REPO_ROOT, tmpFile) returned the absolute path unchanged (no relative traversal is representable across drives) rather than the relative form the test assumed. The remote gsd-test runner this repo gates pushes on is Linux-only and could never have caught this; GitHub CI's windows-latest job is the only signal that does, and it did. Fixed the helper itself (scripts/lint-vendored-deps.cjs's new resolvePath()) to treat an already-absolute input as absolute-in, absolute-out instead of silently mis-joining it, and updated the test to pass the scratch file's absolute path directly rather than relying on a relative conversion that is not always representable. Kept every mutation-sensor assertion intact and added coverage proving resolvePath is a no-op for relative inputs and correctly passes absolute ones through unchanged. Refs #3881 * fix(#3881): warn when state sync regenerates over unparseable frontmatter state sync (ADR-3408 §8.3's sanctioned regenerate path) correctly overwrites an unparseable frontmatter block per its 'body wins' contract — that overwrite behavior is unchanged here. The defect was the silence: synced:true/exit 0 gave no signal that the existing block (including git merge-conflict markers) could not be parsed and was destroyed, per ADR-3473 §8.5 ('a derived conclusion may not be reported as authoritative when the derivation dropped input it could not resolve') and §8.4 ('failure is a value'). Adds a gsd: warning — ... (#3881) line on stderr, matching the existing #3573 precedent, and surfaces the same disclosure in the JSON result's existing changes[] array so a machine consumer sees it too. Exit code and synced:true are left unchanged — sync did what its contract says. REGENERATE_STATE (/gsd-health --repair's sibling on the same sanctioned-regenerate list) is DESTRUCTIVE-risk and unconditionally refused by applyRepairs's dispatcher before runRepairAction ever runs (src/health-diagnostic.cts), so it is not a live path today and is not in scope for this fix. Refs #3881 * fix(#3881): exit non-zero when a state command returns an error Refs #3881 * chore(#3881): changeset for the state exit-code fix Refs #3881 * fix(#3881): honor the documented --project-dir flag Refs #3881 * revert(#3881): restore exit-0 result envelopes for state errors Reverts 9638f2936 and its changeset. The change was wrong and the revert is the correction. This repo distinguishes two error mechanisms deliberately. error() in src/io.cts writes to stderr and calls process.exit(1) -- the hard-failure path. output({error: ...}) writes a JSON result envelope to stdout and returns normally with exit 0. The reverted commit converted 23 result-envelope sites into hard failures, which is a different contract, not a bug fix. tests/state-contract.test.cjs's errorPathDoesNotPublish asserts the envelope contract directly -- a failing command exits 0 with a JSON error envelope and must not publish state.json -- and the remote matrix run caught it along with four cases in the QA scenario walk. Thirteen tests in tests/state.test.cjs that the original commit rewrote were encoding that real contract, not the bug it claimed; they are restored. Whether an error envelope on stdout with exit 0 is the right CLI design is a genuine question, and it is section 8.4's rule ('failure is a value') with its own phase. It is not something to flip inside this PR. Refs #3881 * chore(#3881): backfill changeset PR number for the project-dir fix Refs #3881 * test(#3881): keep the frontmatter mutation shard inside its time budget The Stryker (frontmatter) shard hit the documented 15-minute (900s) shard cap. Root cause is NOT row-level spawn overhead (contrast the #2790/ core-utils precedent): the three shard test files' own logic runs in ~413ms total (356+30+27ms) with all 392 assertions passing. Instead, src/frontmatter.cts grew from ~825 to 1496 lines (+671/-187) migrating to the vendored YAML parser, proportionally growing the mutant count Stryker generates for gsd-core/bin/lib/frontmatter.cjs. Stryker's command runner bills the full 'node --test <3 files>' invocation once per mutant, and node:test's default per-file process isolation forks a child process for each of the three files on every one of those invocations — pure fork overhead multiplied by a much larger mutant population. Fix: scripts/mutation-matrix.cjs COVERED.frontmatter now declares isolation: 'none', and .github/workflows/mutation.yml passes --test-isolation=${{ matrix.isolation }} (defaulting to 'process' — i.e. unchanged behavior — for the other 8 shards, which were not individually audited for cross-file state leakage under shared-process execution). Measured locally via node:test's run() API on the exact 3-file set: isolation:'process' took ~593ms vs isolation:'none' ~478ms for the same 392 passing assertions. The true CI-shard number can only be confirmed on the GitHub Actions run (Stryker cannot run locally, and 'node --test' is hard-blocked in this environment). Refs #3881 * test(#3881): register the vendored-parser tests in the frontmatter mutation shard stryker.config.mjs's own rule ("Keep this list in sync with the tests arrays in scripts/mutation-matrix.cjs COVERED") was violated: #3881 grew src/frontmatter.cts from ~825 to 1496 lines but its new tests (tests/feat-3881-yaml-parser-consequences.test.cjs, tests/frontmatter-golden-parity.test.cjs, tests/frontmatter-roundtrip.property.test.cjs, and +167 lines in tests/frontmatter.test.cjs) were never added to the frontmatter shard's tests array, so Stryker's mutants in the new vendored-js-yaml adapter had nothing constraining them. PR #3888 measured 55.8% against the 65 floor (748 killed / 593 survived / 17 timeout) and the shard was separately cancelled at 15m04s against the 15-minute per-shard cap. Registers all four files (each earns its slot on evidence of a unique constraining assertion, documented inline), gives the shard a measured/projected 180-minute budget via a new per-module timeoutMinutes field threaded through mutation.yml's job-level timeout-minutes the same way isolation is threaded, and removes the prior isolation:'none' override (re-measured at this file-set size, its savings are within run-to-run noise, not worth the unaudited cross-file-state-leakage risk). Refs #3881 * feat(#3881): derive the mutation test list and ratchet the score floor Refs #3881 * test(#3881): ratchet five stale mutation floors and close the frontmatter gap Raised five module minScore floors per CI run 33012034388 (floor(achieved)-1): config-schema 75.51%->74, prompt-budget 88.95%->87, context-composer 79.92%->78, context-utilization 92.31%->91, active-workstream-store 87.42%->86. Updated both scripts/mutation-matrix.cjs COVERED entries and tests/mutation-matrix-ratchet.test.cjs RATCHET_BASELINE in the same diff per the ratchet's own contract. Closed the frontmatter shard's 63.03%-vs-65 gap with new behavioral tests in tests/feat-3881-yaml-parser-consequences.test.cjs, each paired with a documented near-miss: frontmatterDeepEqual's array-order/length/type-mismatch/key-order semantics (via spliceFrontmatter's no-op guard), scalarNeedsDoubleQuoting's leading/trailing-whitespace and dash/surrogate triggers (via reconstructFrontmatter), repairAmbiguousColonValues' already-quoted vs ambiguous-colon repair paths (via extractFrontmatter), and the null-byte sentinel round-trip surviving at region offset 1. Did not lower minScore. Refs #3881 * test(#3881): decouple the ratchet test from real module floors The CLI end-to-end rows in tests/mutation-score-ratchet.test.cjs hardcoded config-schema's real floor (52), which commit 973321541 legitimately ratcheted to 74 -- breaking a test pinned to the exact value the mechanism under test exists to change. Add an injectable --matrix seam to scripts/check-mutation-score-ratchet.cjs and point the CLI rows at a synthetic module + synthetic floor built via a temp fixture, so the rows are indifferent to any real module's floor moving while still exercising the same fail/pass behaviour. Refs #3881 * refactor(#3881): parse must_haves with the vendored parser and drop re-implemented leniency Refs #3881 * fix(#3881): restore the ambiguous-colon repair its hand-edited-STATE.md contract needs A tracked-document sweep of 910 *.md files cannot see this dependent: repairAmbiguousColonValues's one real caller is user hand-edited STATE.md content that never lives in this repo's tree, only on end users' machines, and is pinned by tests/smart-entry.unit.test.cjs. Restores the function plus its post-throw fallback path (loadWithAmbiguousColonRepair) only; repairMalformedInlineArrays and splitLegacyInlineArrayItems stay deleted, reverified against the full frontmatter test shard. Adds a frontmatter-level regression row in tests/feat-3881-yaml-parser-consequences.test.cjs so the dependency is visible where the function lives. Closes #2571 Refs #3881 --------- Co-authored-by: sim <sim@local>	2026-08-26 19:29:32 -04:00
Tom Boucher	c1f755d75d	docs(#3889): ADR-3889 — the process-exit contract, nothing fails with success (#3890) ... * docs(#3889): ADR-3889 — the process-exit contract Records the decision for epic #3889: one Outcome vocabulary, one pure projection to an integer, two terminators sharing both. Doc-only. Regenerates docs/adr/README.md via gen-adr-index. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(#3889): rewrite ADR-3889 as an exit-code registry, on measured evidence The first draft's census was grep-derived and roughly 2x inflated: it counted comment prose and also matched process.exitCode, the correct pattern. An AST census (@typescript-eslint/parser, CallExpression on process.exit) gives 128 deduped call sites, 71% of them in hooks/, against 88 process.exitCode assignments already in use. scripts/ is effectively migrated (39:1). Corrects the framing accordingly: the seam (src/cli-exit.cts) already exists and is already adopted; what is missing is an allocator and a hooks adapter. Replaces the fixed six-value enum with a registry: 0 and 1 free, 2 reserved to the Claude Code hook protocol, 3-13 forbidden (Node reserves them), 64+ allocated with one meaning and one owner per code. Domain codes permitted. Adds three findings obtained by reading and executing the code: - ui-safety-gate/api-coverage report empty stdin as an authoritative negative verdict (exit 1) — reproduced with a control - four modules hand-roll the same three-outcome convention, documented only in a comment at src/ui-safety-gate.cts:137 - two capability fragments fabricate {"detected":false} on probe failure, while the honest {"skipped":true,"reason":"sdk-failed"} form already exists in-tree Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-26 16:53:38 -04:00
Tom Boucher	389bc86e02	enhance(#3883): route every slug call site through its canonical owner (#3896) ... * docs(#3883): correct section 8.3 against the tree I wrote this section in Phase 0 stating rules I had not executed against the tree. Measured at 832dcbb75, three statements are wrong. The slug count is 11 across 5 files, not 13; two of the thirteen were an unrelated tokenizer regex. The divergence is real and reproduced. resolveRuntime reads no install marker at all and has no cache; PR #3382, cited as prior art for that rung, is closed unmerged. The Codex sandbox is still a hand-maintained subset map with a silent read-only fallback, and validate agents checks file presence only -- both halves of that claim are false. The shortFormToId rule is accurate. The guard roster names no casualty for this rule. Section 8.3 is therefore a work list, not a conformance check. Refs #3883 * test(#3883): failing-first rows for slug re-implementation divergence Refs #3883 * feat(#3883): route every slug call site through its canonical owner Migrated commands.cts:209, init.cts:176/1935/1957/3109, phase-id.cts:229/380, phase-locator.cts:269, workstream-name-policy.cts:75 to generateSlugInternal (core-utils.cts:107). Declared different: gsd2-import.cts:97 (no truncation), active-workstream-store.cts:97 (fixed ASCII env-key domain, already matches). Fixed a latent circular-require bug: core-utils.cts top-level destructured comparePhaseNum/scopeToPhase from phase-id.cjs; switched both sides of the new circular require to lazy function-body requires. Refs #3883 * fix(#3883): restore per-site truncation contracts broken by the slug consolidation Refs #3883 * fix(#3883): close review findings — changeset, miscounts, loose rows, cyclic destructure Refs #3883 * docs(#3883): correct the same slug miscount in the Context table Section 8.3's rule text was corrected to 11 copies across 7 files; the Context table above it still carried the original 13 across 5. Same wrong claim, second location, both mine. Refs #3883 * chore(#3883): backfill changeset PR number Refs #3883 * test(#3883): keep the core-utils mutation shard inside its time budget PR #3896's Stryker (core-utils) shard was cancelled at the 15-minute shard cap. Stryker's command-runner bills one `node --test <file>` invocation as a single unit costing whatever the file's slowest run costs, re-run once per mutant (documented in tests/state-contract.test.cjs's header, #2790 precedent). A3/A5 drove every CLI-reachable slug site through runGsdTools (a real child-process spawn per call, ~85-170ms each across ~70 calls), accounting for ~7.1s of the file's ~7.9s wall time. commands.cts:cmdGenerateSlug and init.cts:cmdInitExecutePhase/ cmdInitPhaseOp/cmdInitProgress are plain functions reachable in-process from the built gsd-core/bin/lib/*.cjs, so this calls them directly instead of spawning gsd-tools, capturing their fd-1 JSON output with the bug #1008 fs.writeSync-mock pattern already used in tests/io.test.cjs and tests/init.test.cjs. A hermetic-env helper reproduces the isolation runGsdTools's { HOME: tmpDir } + testEnvBase() gave the child process. File wall time drops from ~7.97s to ~1.1s (246/246 passing, same coverage), well inside the 15-minute cap even at hundreds of mutants. Refs #3883 --------- Co-authored-by: sim <sim@local>	2026-08-26 16:03:11 -04:00
Tom Boucher	a638ca4332	enhance(#3882): stop sentinel phases skewing estimation calibration (#3893) ... * test(#3882): failing-first rows for sentinel phases skewing calibration Adds A1a/A1b/A2/A3 to tests/estimate-calibrate.test.cjs, the module's existing test file, rather than a new bug-NNNN file. collectCalibrationSamples (src/estimate-cli.cts:206) does a raw readdirSync over .planning/phases and never applies isSentinelPhaseId, so a sentinel phase (milestone 0 or 999) carrying a PLAN estimate / SUMMARY actuals pair contributes a phantom calibration sample. computeCalibration is median-based, so a single 50x outlier among three samples leaves the factor unmoved — asserting "the factor is unchanged" against one sentinel would pass on the broken code for the wrong reason. Each row instead asserts the WHOLE computed CalibrationResult object (factor, applied, confidence, sampleCount, clamped) for a sentinel-free project against its sentinel-injected twin: - A1a: one sentinel flips applied false->true and confidence low->med on phantom evidence (calibration switches on with zero real signal). - A1b: two sentinels corrupt the factor itself (1 -> 3, clamped false->true). - A2: the sentinel's own sample is verified absent from the returned list. - A3: the two genuine phases still contribute their own unchanged samples (regression pin — stops A1/A2 passing by filtering everything). Verified RED on today's code (node tests/estimate-calibrate.test.cjs): A1a/A1b/A2 fail with the exact differing objects; A3 and all pre-existing rows in the file remain green (no collateral). Refs #3882 * feat(#3882): route phase enumeration through its owner and name the sentinel axis Task 1: collectCalibrationSamples (src/estimate-cli.cts) hand-rolled a raw readdirSync over .planning/phases, treating every directory (including sentinel phases, milestone 0/999) as a completed phase and feeding phantom PLAN/SUMMARY samples into the estimation calibration factor. Routed through the existing owner, listMilestonePhaseDirs(phasesRoot) with no cwd -- already 'all milestones, sentinels excluded', exactly the combination this caller needs; no new API was required for this half. It now also surfaces the scope discriminator: an unreadable phases directory throws PhasesUnreadableError instead of silently returning zero samples, and cmdEstimateCalibrate reports it via a new ERROR_REASON.ESTIMATE_PHASES_UNREADABLE instead of persisting a phantom empty calibration document. Task 2: added listAllPhaseDirs(phasesDir, { includeSentinels }) to src/phase-locator.cts -- the one genuinely missing axis: 'physical set, sentinels INCLUDED'. includeSentinels has no default and is required, so a call site cannot obtain sentinel-inclusion by omission (compile-time refusal, not just documentation). Mirrors listMilestonePhaseDirs's absent/unreadable scope handling. Task 3: migrated the two exemptions whose written reason maps cleanly onto 'physical set, sentinels included' -- cmdRoadmapAnalyze's _phaseDirNames (src/roadmap.cts) and cmdInitMilestoneOp's diskPhaseDirs (src/init.cts), both heading->directory lookup indexes. Left the rest: archivePhaseDirectories's own body has no readdirSync to migrate (its callers already resolve dirs before calling it, and both current callers deliberately EXCLUDE sentinels -- migrating it would be an unauthorized behavior change, not an API swap); cmdValidateHealth's exemption is vestigial (its actual physical-set sweep already lives in planning-snapshot.cts's buildAllPhaseDirNamesField, a pre-existing near-duplicate of the new axis, flagged as a finding, not restructured); cmdPhasesClear/cmdMilestoneComplete/cmdVerifySchemaDrift/detectHasPriorPhases/detectUiPhaseActive want a different combination (sentinels excluded, or a single-phase lookup) and are unaffected. Task 4: detector 2 (sentinel literal) is untouched and retained. Removed exemption entries only for the two migrated call sites; every other function-scoped exemption is preserved. Guard exits 0. Refs #3882 * refactor(#3882): delegate the snapshot phase-dir scan to its owner buildAllPhaseDirNamesField duplicated listAllPhaseDirs's own readdirSync + directory-filter + absent/unreadable handling — the 'one implementation per rule' defect ADR-3473 SS8.3 names, introduced by this branch's own #3882 work. Delegate to listAllPhaseDirs and re-apply the field's existing lexicographic sort on top, since W007's observable order must not change. Refs #3882 * docs(#3882): document the sentinel axis and the enumeration consolidation Records listAllPhaseDirs in the Phase Locator glossary entry, and the fact that the owner already answers the all-milestones sentinel-free question when called without a cwd -- the call collectCalibrationSamples was missing. Also notes that buildAllPhaseDirNamesField now delegates rather than carrying a second readdir, and that exactly one readdirSync over the phases directory remains across the two modules. Refs #3882 * test(#3882): close review findings — real order proof, unreadable coverage, collision fixtures Refs #3882 * chore(#3882): backfill changeset PR number Refs #3882 --------- Co-authored-by: sim <sim@local>	2026-08-26 15:03:12 -04:00
Tom Boucher	ddde001af6	enhance(#3873): the STATE.md schema — one owner, generated artifacts (#3880) ... * test(#3873): failing-first locale parity, plus tripwires for what must not move Pins ADR-3473 §8.8 at the artifact a reader actually sees. The English STATE.md reference carries a Status lifecycle section that is missing from all four translations — the section documenting the status enum whose clobbering is #3853. The test derives the heading set rather than hard-coding the missing one, and names the locale and the heading when it fails. Two tripwires that must pass today and after. The field-drift guard still catches a re-derived fallback ladder: §8.8 instructs deleting that script, and that instruction rests on a wrong premise about what it guards, so the test stops a future reader from deleting it on the ADR's word. And last_activity's label resolution is pinned to what ships today, because it is declared in one of the two tables this phase consolidates and not the other — the consolidation must not silently pick a side. The locale test buckets under docs rather than state, which is what it tests; that bucket is allowlisted with justification rather than folded into an unrelated docs suite. It reads only markdown, so it carries no allow-test-rule marker — a marker there would suppress nothing and would grow the unverified pool against its ceiling. Refs #3873 * feat(#3873): one schema owns the STATE.md key set, three tables become projections ADR-3473 §8.8. The key set was declared in four places that had to agree by hand and already did not: FIELD_CLASSIFICATION, FRONTMATTER_BODY_SOURCE, FRONTMATTER_KEY_TO_BODY_LABEL and buildStateFrontmatter's emit behavior. One frozen null-prototype schema now declares each key's type, enum, cardinality, source, preservation, body source, body label, accepted parse shapes and whether it is emitted unconditionally; the three tables are derived from it at module load. The projections are byte-identical to the literals they replace, key order included, and the parity tests compare against verbatim copies of today's tables rather than re-deriving both sides from the schema — a parity test fed from one source proves nothing, which is how a consolidation ships a changed policy under a green test. last_activity was the live disagreement: present in one table, absent from the other. The schema declares what ships today rather than the tidier answer, and a test pins it. The schema is a leaf module and owns the four field-policy types, re-exported from state-transition so existing importers are untouched — the same split health-diagnostic-types made to break a CJS require cycle. Refs #3873 * feat(#3873): generate the schema-derived regions, parity-check the prose tables ADR-3473 §8.8's generator half. gen-state-md-docs.cjs owns marked regions in the shipped template and all five reference docs, follows gen-features.cjs's fail-closed contract, and is wired into regen:derived and lint:generated-sync. The Status lifecycle section was missing from all four translations — the section documenting the status enum behind #3853 — and is now generated into every locale. Field cardinality is a new generated table: pure schema data, no prose, so nothing to lose. The Field-reference and Status-values tables are parity-CHECKED rather than generated. Their Purpose, When-populated and Matched-text columns are genuinely hand-translated per locale, and §8.8 itself says prose stays hand-translated; generating them from an English registry would overwrite four locales' translations on every write. The row set is checked against the schema instead, so a key added to one and not the other fails, which is what field drift actually means. Building that check found last_activity_desc undocumented in all five tables. Three keys the docs describe are absent from the schema — active_phase, next_action, next_phases. They are grandfathered by name, not by wildcard, so a fourth fails: a declared gap with a forcing function rather than a silent one. Refs #3873 * fix(#3873): declare what the parsers do, and close the shape-parity gap Two declarations in the new schema described intended behavior rather than actual — the defect class this epic exists to end, committed inside the epic. Both were caught by executing the parsers instead of reading their docstrings. current_plan.acceptedShapes claimed ['N', 'N of M']. Standalone, the hybrid shape errors; the path that looks like support is parseInt truncating '2 of 5' to 2 and discarding the rest. Narrowed to ['N']. The parser is deliberately NOT fixed here: that is #3784 and PR #3791 is already doing it. When #3791 lands this row must widen, and the shape test will go red until it does — the schema and the parser cannot drift apart quietly, which is what §8.8's checked-not- generated rule is for. STATUS_LIFECYCLE_ENUM claimed to be the closed set status can hold. normalizeStateStatus passes unrecognized prose through unchanged, so it is not closed at runtime. The seven members are the canonical values it maps onto; the docstring now says that and the test asserts the real lenient contract. Closes the acceptance item that a test asserts the parsers accept exactly the declared shapes: the check is table-driven over every row carrying acceptedShapes, guarded against passing vacuously on an empty set, and fails loudly if a future row has no registered driver. Adds the unwired-label throw and the fast-check property that every projection agrees with its schema row. Refs #3873 * fix(#3873): keep the shipped template's frontmatter first, and make row 27 able to fail The remote matrix caught 12 failures with one cause. Making the template's frontmatter a generated region wrapped it in its own yaml fence ahead of the markdown fence, so extractFileTemplate and readShippedStateTemplateBody — which both match the single markdown block — found the heading first, not the frontmatter. That breaks the contract every new project's STATE.md is created from: bug #21 and epic #1969 B8 pin that the File Template block starts with frontmatter and carries gsd_state_version. The markers now sit inside the single markdown fence, so the fence opens before the frontmatter and the region still ends ahead of the heading. Same layout as before this phase, with markers embedded rather than a second fence. Row 27 existed to catch exactly this and did not, because it was writer-seeded: it asserted against the generator's own output shape, so it passed on the broken template. It now parses the fence the way production does and was verified to fail against the broken shape before being trusted against the fixed one. A test that would not have caught the bug it exists to prevent is worse than no test. The emitted-attribution failure was separate and the fragment was the wrong remedy: gsd-core/templates/state.md self-attributes under a verbatim-copy identity rule, so a diff touching it needs no acknowledgment. Fragment deleted rather than left explaining nothing. Refs #3873 * docs(#3873): how to change the STATE.md schema The phase gate was right and my docs artifact was wrong. I listed lint:generated-sync as the second enablement step, which is a verification command dressed as one, and then claimed a one-step sequence owed no how-to. The real sequence is build:lib then regen:derived, and the ordering is a trap: the generator reads the COMPILED schema, so regenerating before building regenerates against the previous schema and commits artifacts that look plausible while disagreeing with the code just written. A reference table cannot carry an ordering dependency; that is what the how-to test is for. The page covers adding, changing and removing a key, every reason code the check emits and what to do about each, what is generated versus hand-translated and why the two prose-bearing tables are parity-checked instead of generated, adding a language, and the three grandfathered keys. Indexed from docs/README.md. Refs #3873 * chore(#3873): backfill changeset PR number --------- Co-authored-by: sim <sim@local>	2026-08-26 01:57:47 -04:00
Tom Boucher	3b18eff388	enhance(#3872): what a command reports it wrote — the transaction diff (#3878) ... * test(#3872): failing-first regressions for what a command reports it wrote Pins ADR-3473 §8.7 at the consumer's output. state planned-phase advances current_phase on disk and never reports it, and reports progress.total_plans which reconcileReportedFields silently drops because it cannot resolve a dotted key against nested frontmatter. Both directions of #3818's own before/after diff, reproduced against the real CLI. Also pins the two properties the change must not break: a fully-failed patch still reports an empty updated array, which is what state.cts:607's success boolean depends on; and two content-identical writes differ in last_updated alone. That second one measured state_head NOT to be ambient — it is recomputed every write but only changes when git HEAD moved — so the provenance exclusion is a one-element set, with a companion test pinning that state_head does change when HEAD moves. Refs #3872 * feat(#3872): derive what a command reports from the transaction diff ADR-3473 §8.7. reconcileReportedFields compared the transform's own output against persisted bytes and then filtered what preservation had restored by its FIELD_CLASSIFICATION policy. Both are replaced by one comparison of persisted against the pre-write state the transaction already holds, surfaced to the command through the same caller-allocates out-param idiom divergedFields established. Both of the old directions fall out of that single comparison: a field the transform reported but the pipeline discarded is persisted-equals-snapshot and drops out, and a field nobody reported but the write moved is different and appears. The classification filter is deleted, not relocated — no policy test remains anywhere in the reporting path. Reporting is at dotted-leaf granularity, enumerated from the progress.* rows FIELD_CLASSIFICATION already declares rather than by walking user data to arbitrary depth. That closes a live defect: plannedPhaseCore already pushed progress.total_plans and reconcileReportedFields silently dropped it, because a flat hasOwnProperty cannot resolve a dotted key against nested frontmatter. Current Position was lost the same way and is fixed in the same place. The exclusion is one field, last_updated, and it is by provenance rather than by classification: it is the only field measured to change on every write regardless of content. state_head was measured NOT to qualify — it is recomputed every write but only changes when git HEAD moved. Without that exclusion state.patch's success boolean, which is updated.length > 0, would be permanently true and a fully-failed patch would report success. Refs #3872 * fix(#3872): cover the matrix, and close a prototype-chain read the coverage found Review found 20 of 29 test-matrix rows uncovered. Covering them found two real defects rather than merely documenting the intended behavior. bodyLabelFor read FRONTMATTER_KEY_TO_BODY_LABEL with a bare bracket index on a plain object literal, so a field named __proto__, constructor or toString resolved to the inherited prototype member and leaked a non-string value into the updated array. Fixed with an own-property check, mirroring the discipline resolveFrontmatterPath already had. The security-relevant matrix row proved it before the fix. applyPostSyncPreservation still carried its own inline copy of the value comparison alongside the new stateFieldValuesDiffer, which is two live copies of one rule introduced by the epic that exists to remove them. Routed through the single owner. Adds the fast-check property that a field appears iff its persisted value differs from the snapshot, the string-versus-number representation boundary, dotted paths into missing parents and into scalars, deleted and added keys, and the preserve-if-placeholder pair that proves no classification test survives in the reporting path. Refs #3872 * docs(#3872): document the transaction diff on the write path The updated array's contract belongs where the write path is described. States the iff rule, leaf granularity, the single provenance exclusion and why state_head is deliberately not one, and closes with the consequence a reader actually needs: these arrays are longer than they used to be, because they used to under-report. Refs #3872 * fix(#3872): a derived leaf materializing is not a change the caller made The remote matrix caught 17 failures with two causes. The substantive one is that progress is source: disk, and the disk cannot change during a STATE.md write — the write only touches STATE.md. So a progress block appearing where the snapshot had none is the scanner populating a document that had never been synced. The bytes moved; nothing the caller did moved them. That is the same shape as last_updated one level up, so the provenance rule is generalized rather than special-cased: a field appears iff its persisted value changed for a reason attributable to this write's action, and two cases are not attributable — a field stamped unconditionally on every save, and a declared derived leaf materializing from a source that did not change. Crucially this does not consult the preservation policy, so the filter §8.7 deleted stays deleted; it uses the declared leaf set to know which keys are derived. This had a second production consumer the earlier review concluded did not exist: cmdStatePlannedPhase gates publishStateContract on updated.length, and its own inline comment predicts exactly this failure. A no-op call was publishing state.json. advancePlanNoOpDoesNotPublish genuinely encoded pre-§8.7 behavior and moves. E2 and E6 had carved out total_plans as reportable-on-materialization, an error introduced earlier on this branch rather than a pre-existing pin, and are corrected with it. Refs #3872 * chore(#3872): backfill changeset PR number --------- Co-authored-by: sim <sim@local>	2026-08-25 23:51:16 -04:00
Tom Boucher	8f674281fd	chore(#3875): sweep the spent ack fragments and automate the sweep (#3877) ... * chore(#3875): sweep the spent ack fragments and automate the sweep next has been red on every push since a84f75630 (#3823) — 24 consecutive pushes over two days — on two fully-spent emitted-drift-ack fragments nobody swept. #3823 introduced guard-no-ack-on-next together with a 45-fragment sweep, but computed that sweep as a static set of deletions fixed at its branch point. #3809's fragment merged to next while #3823 was in flight, so the guard reds on its own merge commit. The condition is evaluated dynamically at merge time and remediated statically at branch time; on a moving branch the second can never reliably satisfy the first. - delete tests/emitted-drift-acks/3809-* and 3866-* (3034-* and 3172-* stay -- the #3842 open-PR hold correctly defers them) - runGuardNext returns `sweepable`, the set the guard actually reasoned about, plus `legacyPresent` for the legacy document, which is a fixed path rather than a fragment basename and would otherwise be invisible to any sweeper - new --sweep-plan mode turns the guard into a work list: plan on stdout, prose on stderr, exit 0 so a non-empty plan does not fail the step that asked for it - main() is injectable in BOTH lanes; a half-injected seam lets a test that passes cwd silently read the real repository instead of its fixture - ack-fragment-sweep.yml derives its deletion list from that plan on a timer and opens a reviewable PR, so the sweep can no longer go stale between branch and merge Hardening found in review, each verified against a live reproduction: - git rm reads its arguments as PATHSPECS with wildmatch semantics, so a fragment named a bare-star .json name -- legal, and admitted by listFragmentFiles since it filters only on the suffix -- expanded to every fragment in the directory, including ones the #3842 hold withheld. Confirmed in a scratch repo: one such file deleted all three. Closed with a literal allowlist and a :(literal) pathspec, two independent layers. - an apostrophe inside a heredoc nested in a command substitution is an unterminated quote and a hard syntax error at runtime, not just under bash -n. - an empty plan no longer reports success unconditionally: the guard is re-run without the hold to tell "next is clean" from "everything is held", the commonest holder being the sweep PR from the previous run, which touches exactly the fragments it proposed to delete. - a branch pushed by a run that died before it could open the PR wedged every later run on a non-fast-forward push; re-pointed under a lease instead. - a guard crash in plan mode no longer reads as "nothing to sweep". Refs #3875 * chore(#3875): regenerate CONTEXT-INDEX.json for the glossary entry lint:generated-sync failed on CI: gen-context-index.cjs derives docs/CONTEXT-INDEX.json from CONTEXT.md, and the RULESET.EMITTED_ATTRIBUTION entry added in the previous commit left it stale. Refs #3875 * chore(#3875): regenerate the example CONTEXT-INDEX for the glossary entry CONTEXT.md feeds TWO committed indexes, not one: docs/CONTEXT-INDEX.json via scripts/gen-context-index.cjs, and the examples/dynamic-context-management copy that lint-example-parser-parity holds to a fresh parse. The previous commit regenerated only the first, so the parity check stayed red. Refs #3875 --------- Co-authored-by: sim <sim@local>	2026-08-25 23:17:29 -04:00
Tom Boucher	1863f5569c	enhance(#3871): the state transaction — mandatory snapshot, open()/rebuild() (#3874) ... * test(#3871): failing-first regressions for the dropped curated progress block Pins ADR-3473 §8.6 / #3756 at the consumer's output: state record-session and state add-decision on an archived-milestone project drop the curated progress frontmatter entirely, exit 0, and report nothing. Reproduced against the real CLI before writing the tests, not inferred from the issue text. Also adds the unit-level probe that applyStatePreservation's preserve-always row is inert on a resyncing write, and an over-preservation guard that an empty project is never inflated. Refs #3871 * feat(#3871): make the STATE.md pre-write snapshot mandatory via open()/rebuild() ADR-3473 §8.6. StatePreservationInput's nullable preFm and the always-present preFmSnapshot were the same extractFrontmatter call, one of them nulled on resync — a policy flag baked into a snapshot. Both collapse into a single StateTransaction whose snapshot cannot be absent: openStateTransaction() applies preservation, rebuildStateTransaction() does not, and both carry the snapshot because the reporting phase needs it either way. An absent snapshot is now a construction failure; an empty one stays legal, because that is what a document with no parseable frontmatter honestly has. writeStateMd requires a rebuild transaction, which types ADR-3408 §8.3's closed exception list at both call sites (state sync, health --repair) instead of matching them as strings in a ratcheted baseline. Fixes the dropped curated progress block: an all-zero or absent derived total set is an unmeasured scan, not a measurement, so the curated block stands. Also fixes two defects surfaced while building — preserve-always reported a mutation even when it restored an identical value, and it re-entered the curated object by reference, which would alias the snapshot the next phase diffs against. Refs #3871 * fix(#3871): close the three remaining subsumed defects and restore the arm the type does not replace Review of the first two commits found four things. The guard shrink deleted the seam-bypass axis whole, but only its writeStateMd( arm became redundant. Its other arm catches a call site re-assembling syncStateFrontmatter + applyPostSyncPreservation instead of the owned composition, which the transaction type does not make unrepresentable and which #3469 found live. Restored as findCompositionBypasses, terminal rather than ratcheted. Three of the four issues this phase claims were untouched. All three are the epic's own shape and are fixed at the seam: current_phase_name is reasserted from the curated value when the caller names none, and cmdStateJson stops carrying a hand-maintained list parallel to FIELD_CLASSIFICATION and projects it instead. The construction failure that is the point of this phase had no test. Every enumerated matrix row now has one, including the measured-versus-unmeasured coercion boundary and a seeded property that no curated key is ever dropped. ADR-3473 §8.6 said the guard 'keeps only its raw-write check'. Verified against next: there was no raw-write check, and four other checks it does not name. Amended in place with the evidence. ARCHITECTURE.md separately advertised a preservation policy the code had deleted. Refs #3871 * fix(#3871): do not let the unmeasured-scan rule block an explicitly-requested resync The remote matrix caught over-preservation, the failure this phase's own negative space says must not happen. state update Progress re-derives the block from the body the caller just rewrote; on a project with no phase dirs the derivation yields zero totals, the unmeasured rule read that as 'the scan measured nothing', and the stale curated percent was restored over the resync the user asked for. preserve-always already said what the missing condition was: never overwrite unless the caller explicitly names this field. explicitProgressField carries it and is derived from shouldResyncStateProgress, not set by hand at a call site, so it cannot drift from what the caller asked for. Two defects found in the same mechanism and fixed with it. readModifyWriteStateMd enumerates its option keys, so a new option was silently dropped rather than rejected. And the raw-write axis captured its first argument up to the first comma, which lands inside a nested path.join, so a write to a STATE.md literal was invisible to it — the prove-it-can-fail test caught that one immediately. No test assertion was weakened; all three frontmatter rows encode #3242, #1969 B3 and #1972 and stand unchanged. Refs #3871 * docs(#3871): record why the raw-write check is kept, not why it was named The amendment justified findRawStateWrites as 'written because §8.6 requires it to exist', which is cargo-culting the contract and would have been the wrong reason to keep anything. The real reason is that writeStateMd acquires the STATE.md lockfile and a raw fs.writeFileSync acquires nothing, so this is a lock bypass and lost-update is the #500/#905/#1230 family — and after this phase it is the one reachable path into the file that nothing else covers. Also records why ADR-3408 §8.6's deletion of the 'clear' policy is not the precedent it looks like: 'clear' was dead vocabulary in a closed enum, this is coverage of a reachable path. Refs #3871 * chore(#3871): backfill changeset PR number --------- Co-authored-by: sim <sim@local>	2026-08-25 20:38:12 -04:00
Tom Boucher	382bf7c423	fix(#3706): deliver the resolved reasoning effort to OpenCode subagents (#3867) ... * test(#3706): failing-first coverage for OpenCode variant emission and frontmatter escaping * fix(#3706): emit the resolved reasoning effort as OpenCode's variant key `query resolve-execution` resolved an effort level for every agent, but the OpenCode bake wrote only `model:` — the effort never reached the generated agent, so subagents ran at whatever the runtime defaulted the model to. This is the effort-side twin of the model-side defect fixed in #3705. The key is written only when an `effort` block is actually configured. `resolveInstallTimeEffort` always returns a level (the catalog default is `high`), so gating on its return value would stamp `variant: high` into every existing OpenCode install — and OpenCode resolves a variant name against a `variants` map in the user's `opencode.jsonc`, so a value nobody declared is not a safe default. Gating on `readGsdEffectiveEffortConfig` keeps installs that never asked for effort routing byte-identical. Kilo does not receive the key: `EFFORT_ARGV` declares surfaces for claude, opencode and codex and has no kilo entry. This is deliberately asymmetric with the model side, where #2794 J8 requires the two runtimes to resolve alike. Both frontmatter sinks now route through `frontmatterScalar`, which quotes and escapes any value that is not a plain scalar. The raw interpolation predates this change, but it was already shown by execution during the #3705 security review to let a config value containing a newline inject additional top-level keys (`tools:`, `permission:`) into a generated agent file. This change adds a second write to that sink, so it is closed here rather than doubled. * fix(#3706): quote frontmatter values YAML would not read back verbatim Self-review of the predicate added in the previous commit. Treating /^[A-Za-z0-9._:/@+-]+$/ as 'safe to emit bare' answers the wrong question: a value can match it and still not round-trip. - A leading '@' is a YAML *reserved* indicator and may not open a plain scalar at all, so a scoped ID like '@org/model' emitted bare is a parse error, not an ambiguity — the whole agent file becomes unreadable. - 'no' / 'y' / 'off' / 'null' resolve to booleans and null, so a variant with one of those names would match no entry in the user's variants map. - '12:30' resolves to 750 under YAML 1.1 sexagesimal, and ':' is legal mid-identifier here, so the form is reachable rather than contrived. Real model IDs pass every clause and stay bare, so already-generated files remain byte-identical. * fix(#3706): route variant through the declared effort seam and cover the live path Addresses six findings from the isolated review, all confirmed by execution. The tests were the serious one: they required `../bin/install.js` while the fix landed in src/, which compiles to gsd-core/bin/lib/. They exercised a different copy of the converter than the one the bake actually uses, so the whole suite was green-by-construction against unchanged code and the remote run failed all 13. Every case now runs against BOTH copies from one table, which doubles as the parity assertion the generative-fix note in runtime-artifact-conversion.cts asks for, and bin/install.js carries the mirrored change. Emission no longer hand-rolls the value. It goes through `renderEffortArgv`, the declared OpenCode effort seam (EFFORT_ARGV.opencode: its own supported set and clamp). That is what rejects a level that is not a wire value — above all `inherit`, which per #3533 (10d) means "omit the key and follow the host default" and was previously written literally, naming a variant that cannot resolve. Reachable two ways, both now pinned: an agent_overrides entry and a routing_tier_defaults entry. A bare effort.default does NOT reach a tiered agent (the #3531 tier ladder answers first), so a test written against `default` alone asserts nothing — that is pinned too. The plain-scalar decision moved into frontmatter.cts beside `scalarNeedsDoubleQuoting` rather than sitting next to it as a second, weaker predicate. `agentScalarNeedsDoubleQuoting` is a documented superset: it adds a trailing `:` (read as a nested mapping key, which fails the whole frontmatter), boolean/null words, and numeric-looking values including YAML 1.1 sexagesimal. Docs now state the cascade plainly: the gate is on effort being configured at all, not on the individual agent being named, so every generated OpenCode agent gets a variant line once any effort block exists. * test(#3706): assert the two frontmatterScalar copies cannot diverge A hand-picked adversarial corpus plus a fast-check property over YAML-significant strings, both run against bin/install.js and the live src copy. Verified the property can actually fail: mutating one copy's quoting rule is killed well inside the run budget. * fix(#3706): close the review findings — predicate, seam, and dead mirror Third review round; every item below was confirmed by execution. The scalar predicate was wrong in two families, both found by a round-trip property test rather than by reading. Basing it on scalarNeedsDoubleQuoting dropped the "first character must be alphanumeric" clause, so `~`, `.inf`, `.nan`, `+1`, `-0` and `.5` went out bare and came back as null/floats/ints; and that base predicate only inspects the FIRST character, so an embedded `: ` (a nested mapping, i.e. a parse error) or ` #` (a comment, i.e. silent truncation) also passed. Dates round out the set: `2026-08-25` opens alphanumeric, survives every other clause, and YAML resolves it to a Date. The property now asserts the contract directly over generated values instead of trusting an enumerated character list. The bin/install.js mirror is gone. Its premise was false — install.js already requires bin/lib at :65 — and it was unreachable besides: install.js's convertClaudeToOpencodeFrontmatter has no `isAgent: true` call site, because its agents path resolves converters from the compiled module. It was a third copy of the YAML rules serving a test rather than a caller, so the file is back to origin/next and the tests target the live copy only. Effort clamping moved to `clampEffortForHost`, which renderEffortArgv now delegates to. The layout was calling renderEffortArgv with a hardcoded 'argv' to borrow its clamp, which read as if the frontmatter key were gated on the invocation-time axis. It is not: claude declares effortSurface "argv" and independently bakes an effort: key. One capability table, one clamp, two channels that no longer pretend to be each other. Also corrects an earlier claim of mine: adding EFFORT_RENDERING.opencode would NOT have made `effort sync` write the wrong key, because it guards on the runtime name before it ever renders. The seam choice stands on other grounds. `effort sync` still skips OpenCode, but its stated reason claimed OpenCode "does not use effort: frontmatter", which this change makes false — so the message now says what is actually true. * docs(#3706): restate the changeset around the round-trip contract * fix(#3706): restore the changeset fragment belonging to #3809 An earlier commit in this branch picked the first file in .changeset/ by glob order instead of the fragment created for this issue, and overwrote agile-geese-squeak.md (PR 3815 / #3809) with this change's body. Restored verbatim from origin/next; this change's text now lives in its own patient-cranes-parade.md, where it was created. * feat(#3706): maintain the OpenCode variant key from effort sync Install bakes the resolved effort into OpenCode agent frontmatter as `variant:`, so `effort sync` has to maintain it or a config change only takes effect on reinstall — and its skip message claimed OpenCode does not use frontmatter effort at all, which this issue made false. cmdEffortSyncOpencode mirrors the codex branch: resolve per agent, clamp through the declared OpenCode capability, then write, strip, or skip. A null target means the key must not exist, which covers both "no effort configured" and "resolved to inherit or to an unsupported level" — the same states under which install writes nothing, so sync and install agree by construction. The frontmatter line-editors are key-parameterised rather than copied: setEffortFrontmatter / removeEffortFrontmatter are now thin wrappers over the same internals the variant path uses, and a test pins that the claude `effort:` behavior did not move. The child-process test harness fixes both HOME and USERPROFILE, so the hermetic-config assertions cannot pass vacuously on Windows. * fix(#3706): scope the frontmatter line editors to the matched block Found by the security review of the sync path, reported as correctness rather than vulnerability, and reproduced against pre-fix code before being fixed. Both editors matched the frontmatter with a regex that can match a block after a preamble, then derived the EOL and the opening-fence length from the START OF THE FILE. On a CRLF document with a preamble those disagree, the offsets shift by one byte, and the reassembled document comes back with a mangled fence (`---\rname: x`). Both now take the EOL from the matched block. `setFrontmatterKeyLine` additionally did a whole-file `/m` replace when the key already existed, gated only on the key being present in the frontmatter body — so a preamble line starting with the same key was rewritten instead of the frontmatter one. It now replaces inside the frontmatter span only, which is the hazard `removeFrontmatterKeyLine` already documented and guarded against. Neither is reachable from an install-written `gsd-*.md` (those begin at byte 0 with `---`), and both predate this change — but the editors are in this diff because #3706 key-parameterised them, so they are fixed here rather than left for the next caller to trip over. Three regression tests, each confirmed to fail against the pre-fix build. * fix(#3706): treat a present-but-empty key as present, and pin the real seam Fourth review round. The MAJOR one: both sync branches read the current value with `(.+?)`, which needs at least one character, so a key present with an EMPTY value read as "key absent". When the target was also null the code concluded "already correct" and skipped — leaving the key in the file, where it reads back as YAML `null`: exactly the unresolvable-variant state this change exists to prevent. Whitespace decided whether it fired, since `variant: ` matched and `variant:` did not. Presence and value are now separate questions at both the opencode and the claude branch. The OpenCode writer now follows the codex branch rather than the claude one: tmp file plus retryRenameSync with orphan cleanup, and a write failure skips that agent and is reported instead of aborting the sweep. Same granularity, same transient-Windows-lock exposure, so the hardened sibling was the right precedent. Also: the generic line-editors escape their interpolated key, the JSDoc stranded by the clampEffortForHost extraction is back on renderEffortArgv, and a cast that declared a nullable function as non-nullable is corrected. Tests close the gaps the review listed — empty value (both spellings), CRLF round-trip through write and strip, the symlink guard, a body line starting `variant:`, a file with no frontmatter, and the YAML classes that actually broke the predicate. The new layout-seam test drives the real stage() path and was verified to FAIL when `variant` is removed from the converter call; a seam test that survives cutting the seam is worse than none. * fix(#3706): clear the round-five review findings No blockers or majors this round; the repo's review gate is zero-tolerance, so the minors are cleared too. A duplicated key was only half-stripped: the strip regex had no `g` flag, so a frontmatter carrying the key twice lost one occurrence, reported success, and left the "a null target means the key must not exist" invariant false on disk — converging only on a second run. Such a document is already invalid YAML, so this is robustness rather than a live corruption path, but a successful sync has to leave the invariant true. A run in which every write failed still summarised as `ok`, so a caller could not tell "nothing to do" from "everything failed". The OpenCode branch now reports `failed` when any write failed. The write-failure path was also the newest code in the change with no coverage at all; it now has a test that injects the failure by monkeypatching the write, per CLAUDE.md §4, rather than by chmod — mode bits do not bite under root in CI. `CodexEffortSyncWriteFailure` is renamed `EffortSyncWriteFailure` now that two branches share it. Removed a guard on the claude concrete path that was provably unreachable — no member of EFFORT_SET renders null there, so it read as protection that did not exist. The claude inherit path's presence check is load-bearing and untouched. Three stale statements corrected: the OpenCode result shape matches codex's, not claude's, now that it emits write_failures; the `thread()` test helper now calls `clampEffortForHost` so it genuinely mirrors the layout instead of merely claiming to; and a test helper restored `USERPROFILE` by assignment, writing the literal string "undefined" into the environment on POSIX — it deletes now. * fix(#3706): converge the set path, degrade on unreadable files, preserve mode Rounds five and six of review. No blockers or majors; the review gate is zero-tolerance, so the minors are cleared too. `setFrontmatterKeyLine` was the mirror of a defect already fixed in its sibling: `remove` was made global, `set` was not, so on a frontmatter carrying the key twice it rewrote the first and left a stale second. Last-wins YAML readers honour the stale value while the sync's own first-occurrence read reports "in sync" — permanently non-converging. It now collapses to exactly one occurrence, in the position of the first, so ordinary single-occurrence documents stay byte-identical (verified across seven shapes before and after). An unreadable agent file used to throw and abort the entire sweep, while a failed WRITE in the same loop degraded into a report. The OpenCode branch now reports read failures alongside write failures; the claude branch degrades to a skip without a new result field, because its shape is long-standing and widely consumed and one bad file aborting the sweep is the actual defect. The tmp+rename publish dropped the original file's mode — a plain writeFileSync preserves it, a rename does not — so a 0600 agent came back 0644. Both the OpenCode and the codex branch now carry the original's permission bits across the publish, masked with 0o7777: the raw stat mode includes the file-type bits, and POSIX leaves those unspecified for chmod. Linux is the only OS the remote matrix runs, so relying on Darwin's tolerance would have been untestable here. Also documents the `from` contract on EffortSyncChange (null means the key was absent, '' means present with an empty value — a distinction earlier rounds introduced and then collapsed in the output), adds OpenCode to the docs paragraph enumerating where the key is omitted under inherit, and records in a comment that the 'failed' summary reaches only raw mode and does not change the exit code, which is a CLI-contract change affecting all three branches and is deliberately not made here. * fix(#3706): guard the codex read, close the tmp permission window, rename the failure type Round seven, plus one thing I found myself. `cmdEffortSyncCodex` still had an unguarded `fs.readFileSync` — a read fault on one agent exited 1 and aborted the whole sweep. The claude and opencode branches were both guarded earlier this round and codex was missed, with the unguarded read sitting ten lines above the chmod block the previous commit did edit. It now reports read failures the way the OpenCode branch does, and a read failure flips its summary to `failed` — which write failures did not do there either, so both are corrected for consistency. The tmp file was created at the default mode and only tightened afterwards, so a 0600 agent's contents sat in a 0644 file for the length of the publish. I measured the window rather than assuming it, then closed it by passing the mode at creation. The chmod after the write is deliberately RETAINED and commented: the `mode` option only applies when the file is actually created, so a leftover tmp from an earlier crashed run would be truncated and reused at its old mode, and the chmod is what corrects that. `EffortSyncWriteFailure` is renamed `EffortSyncFileFailure` — it was typing a `read_failures` array, the same naming-lie the `Codex…` prefix had last round. Also pins the codex mode preservation with a test. It only writes on a path that genuinely rewrites the file, so the fixture is an Anthropic-flavoured model pin the sync strips, and the test asserts the content changed before checking the mode — otherwise it would pass on a sync that did nothing. * fix(#3706): guard the claude writes and share one escaping rule The security sign-off caught a comment of mine that was factually wrong: the new claude read guard said the failure is folded in "like the write path in this same loop does", and there was no write guard in that loop. Rather than correct the sentence, both claude write sites are now guarded the way the read is — a failed file is skipped, the sweep continues, and the raw summary token flips to `failed`. The JSON shape stays frozen deliberately, because it is long-standing and widely consumed; the token is the channel that can carry the signal without a compatibility risk, which is the reviewer's own suggestion. That makes all three branches consistent: reads and writes guarded everywhere, per-file failures degrade instead of aborting, and every branch reports `failed` rather than `ok` when something did not sync. `setFrontmatterKeyLine` interpolated its value raw while the install-side writer quoted through the shared helpers — two writers of the same frontmatter key disagreeing on escaping, the divergence class this repo requires closed. They now share one rule. Verified no churn: all six effort levels are plain scalars and emit byte-identically, with claude's documented minimal-to-low clamp the only difference in the table, exactly as before. * fix(#3706): publish claude agent writes atomically too Both reviewers found this independently, and it is data loss rather than a reporting gap. The claude branch wrote in place, so `fs.writeFileSync`'s O_TRUNC meant a post-open fault left the agent file truncated or half-written: an injected ENOSPC produced an empty file, and under `ulimit -f` a 60000-byte agent came back as 512 bytes of wrong content. The guard added earlier this round then counted that destroyed file as `skipped`, which in JSON mode is indistinguishable from "already in sync" — so a caller would have read the sweep as clean while an agent on disk was corrupt. It now publishes the way the codex and opencode branches already do: write to a tmp file created at the original's masked mode, chmod, then retryRenameSync, with the tmp unlinked and the agent skipped on any failure. The corrupting case is gone rather than merely reported, which matters because this branch deliberately takes no new result key. I had claimed all three branches were consistent after the previous commit. That was true for degradation and reporting and not for atomicity; the reviewer caught the overclaim. It is true now. Also sorts the claude file list, which the other two branches already did — readdir order is platform-dependent, so leaving it unsorted made the reported `changes` ordering differ across machines for identical inputs. * chore(#3706): backfill the changeset PR number pr:0 placeholder replaced with the real PR now that gh api returned it. * test(#3706): kill the frontmatter mutants this change introduced CI's Stryker frontmatter shard scored 60.58 against a break floor of 62. The cause is documented in the lane's own config, from #1882: this PR added a multi-clause predicate to frontmatter.cts and exported the escaper, but the tests constraining them live in tests/runtime-converters.test.cjs, which that shard does not run — so every mutant in the new code was uncovered there even though the behaviour is tested elsewhere. The fix is assertions that kill real mutants, per the repo's own instruction, not a lowered floor and not a Stryker disable: scripts/mutation-matrix.cjs is untouched. Each clause of agentScalarNeedsDoubleQuoting now has a true case AND a near-miss that must answer the opposite way, so flipping the clause fails a specific named test — alnum-first against `a-b`, trailing `:` against `foo:bar`, embedded `: ` against `a:b`, embedded ` #` against `a#b`, the word list against `yes1`/`nullish`, the numeric forms against `1a`/`0xzz`, the timestamp against `2026-08-25x`, plus the case-insensitive spellings that pin the `i` flag. escapeDoubleQuoted is pinned on exact output, including a case constructed so that escaping in the wrong ORDER yields a different string. Two of my expectations were wrong and are asserted as the code actually behaves: `12:99` is NOT quoted, because the sexagesimal alternative never range-checks minutes and so does not match — which is right, since YAML would not read it as sexagesimal either; and `20260825` is quoted by the numeric clause rather than the timestamp one, being a bare integer. * chore(#3706): ratchet the frontmatter mutation floor to 65 The lane measured 66.67 on PR 3867 after the mutant-killing unit tests landed — above its pre-change 63.35 baseline, not merely recovered. Step 3 of this file's own HOW TO UPDATE procedure says to set minScore = floor(measured) - 1 in the same diff, so 62 becomes 65 and the improvement is locked in rather than left free to slide back. The ledger of measured scores now records the new measurement, why the shard broke in the first place (logic added to frontmatter.cts whose only tests lived in a file this lane does not run — the same trap the #1882 note describes), and one discrepancy: step 3 also says to update "the matching RATCHET_BASELINE entry", but no such declaration exists in this file. The name appears only in that comment, so minScore and the ledger are all there is to update. * fix(#3706): update RATCHET_BASELINE alongside the raised floor The ratchet test caught the previous commit: it raised COVERED['frontmatter'] .minScore to 65 without updating the baseline that mirrors it, which is exactly the mismatch that guard exists to make visible in review. I had claimed RATCHET_BASELINE did not exist. It does — in tests/mutation-matrix-ratchet.test.cjs, not in scripts/mutation-matrix.cjs, which is the only file I searched before concluding it was a stale reference. The ledger comment is corrected to say where it lives and to record that the guard caught the error rather than leaving my wrong claim on the record. * docs(#3706): put the mutation ledger entries back under their own dates The 2026-08-25 measurement was spliced into the middle of the 2026-06-14 list, so adr-parser, config-schema, active-workstream-store and core-utils ended up sitting under the wrong heading and misattributing their measurement dates. That ledger is what a future change reads to calibrate a floor, so a wrong date there is not cosmetic. Each measurement is now under the date it was taken. Also drops the first-person account of my own mistake from the entry — the factual half (where RATCHET_BASELINE lives, and that it is updated in the same diff) is what a reader needs; the confession is not. --------- Co-authored-by: sim <sim@local>	2026-08-25 19:54:30 -04:00
Tom Boucher	86fa2917d7	enh(#3866): dispatch step and contribution hooks at verify:pre (#3869) ... * test(#3866): pin that verify:pre must dispatch every hook kind verify-work.md's verify_pre_hooks step dispatches only `kind == "gate"`, so getWiredKinds reports verify:pre -> {gate} and gen-capability-registry rejects any capability declaring a step or contribution there. The verify lane is therefore closed to capabilities that want to contribute to what UAT covers rather than refuse to let it start. Failing-first: the step, contribution, and exact-kind-set rows are RED; the pre-existing gate row is a green regression pin so the new arms cannot orphan the arm verify:pre already had. Refs #3866 * feat(#3866): dispatch step and contribution hooks at verify:pre verify_pre_hooks dispatched `kind == "gate"` only, so getWiredKinds reported verify:pre -> {gate} and gen-capability-registry's validateHooksWired rejected any capability declaring a step or contribution there. A capability could refuse to let UAT start; it could not contribute to what UAT covers. Add contribution and step arms mirroring execute:wave:post, deferring to references/loop-hook-dispatch.md and carrying its ref.command in-context validation guard ahead of any shell-use prose. A verify:pre step is advisory: it never blocks the start of UAT and an erroring step is routed by its own onError. The gate arm and its check guard are untouched. Give extract_tests an additive consumption seam for the artefacts those steps declare via the existing steps[].produces field -- no new registry field, no new ordering, no invented filename. Manifest-supplied artefact names are validated in-context against an allowlist and resolved only inside PHASE_DIR. With no producing step the derivation is unchanged, pinned by test rather than asserted in prose. Review findings folded in: the artefact-name allowlist (isolated adversarial pass), the artefact-shape contract and the seam-inertness tests (spec axis), and the reference/how-to split so one constraint has one source of truth (standards axis). Closes #3866 * chore(#3866): backfill changeset PR number --------- Co-authored-by: sim <sim@local>	2026-08-25 18:07:29 -04:00
Tom Boucher	7bbbe495d7	docs(#3473): ADR-3473 — enforcement by construction, one owner per invariant (Phase 0) (#3870) ... Phase 0 of epic #3473 ships this ADR alone. Every rule in §8 carries a status of Enforced or Required — Phase N. §8.1–§8.5 state the epic's B1–B5 criteria as contract rules with no phase assigned yet. §8.6–§8.8 assign Phases 1–3 to the STATE.md family that ADR-3180 and ADR-3408 left between them: the write seam's unvalidated input, its classification-excluded reporting, and the schema transcribed by hand into nine artifacts. Derivation authority is explicitly NOT in scope here — it lands as ADR-3180 Amendment 8, generalizing §7.5's already-locked sentence. Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-25 17:50:34 -04:00
Tom Boucher	aa6c332b5a	fix(#3701): resolve next_phase from the roadmap, selecting the numerically lowest successor (#3852) ... * test(#3701): failing-first coverage for roadmap-order next_phase resolution * fix(#3701): resolve next_phase from roadmap order, keeping the disk scan for spelling and fallback * fix(#3701): select the numerically lowest successor in both scans, not the first row encountered * chore(#3701): backfill changeset PR number --------- Co-authored-by: sim <sim@local>	2026-08-25 08:58:10 -04:00
Tom Boucher	fb2d122d7f	feat(#3841): assert gsd-tools identity on every state-mutating verb (#3848) ... * feat(#3841): assert gsd-tools identity before any state-mutating verb only this package publishes. The path-based branches — a project-local install, a runtime config directory — had no such guarantee; they trusted their configured location. This closes them. Mechanism: once resolution finishes, and before any verb runs, the preamble probes the tool it picked with `runtime-identity --raw` and matches the answer with a shell `case` pattern ANCHORED to the start of the compact payload (`{"packageName":"@opengsd/gsd-core"`). An unanchored substring match accepts the decoy `{"packageName":"get-shit-done-cc","note":"@opengsd/gsd-core"}`, which any colliding package could publish. The outcome is exported as the two-valued `GSD_IDENTITY_STATUS` (`ok`/`unverified`), so the gate is asserted on a VALUE rather than on warning prose. Rollout is warn-then-fail per the #3146 ruling: `unverified` prints one line naming BOTH causes and continues, because `no_identity_verb` cannot tell a foreign package from an `@opengsd/gsd-core` older than the verb, and at rollout the old-version case is the common one. The blocker was byte budget, not design. The preamble is inlined into 112 shipped files and several sat within single-digit bytes of frozen ceilings (`gsd-verifier.md` 16 bytes, `gsd-executor.md` 33, `execute-phase.md` 234); a first attempt broke five of them. What made room was collapsing the resolver's twenty near-identical `elif [ -f … ]` arms into one candidate-list helper (`_gsd_at`), which buys far more than the assertion costs. The preamble is now 2,624 bytes against 4,500 — a net 1,876 bytes SMALLER per inlined file, so every capped file moved away from its ceiling rather than toward it. No cap raised, no size-budget exception added, no override token emitted. Resolution order, every runtime-home probe, the `unset -f gsd_run` re-source fix, the fail-closed `exit 1`, and the `CLAUDE_ENV_FILE` persistence are all preserved byte-for-byte in substring terms; the snippet still begins with `_GSD_SHIM_NAME=` and still ends with `fi`, which the parity extractors anchor on. `gsd-core/references/gsd-run-resolver.md` is re-synced byte-equal. Also fixes two stale claims found in passing: CONTEXT.md and FEATURES.md both described an `[ -x ]` guard as the load-bearing re-source defense. That guard was tried and REMOVED in #3831 — it rejected the bare function name, fell through every branch, and hit `exit 1`, which kills a sourced caller's shell. `unset -f gsd_run` is the actual mechanism. Refs #3841 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3841): pair the anchor's brace by requiring a closed identity payload The matrix went red on `tests/new-project-mvp-prompt.test.cjs` — "new-project.md has unbalanced braces: net depth 2" — plus a knock-on report from its parent `bug #1516` describe, which is the same failure counted once at the child and once at the block. Root cause: that guard (:182-189, mirroring #3784 bd53925f) walks characters and increments on `{`, decrements on `}`, with no awareness of shell quoting. It scans `new-project.md` PLUS every `new-project/steps/*.md`, and both `new-project.md` and `steps/auto-mode-config.md` carry one inlined preamble copy — hence net 2 from a snippet that was off by exactly one. The unpaired brace was the `{` inside the single-quoted `case` pattern of the identity anchor, which is correct shell and invisible to a text scanner. Fix in the snippet, not the guard. The pattern now anchors at BOTH ends: `'{"packageName":"@opengsd/gsd-core"'*'}'`. That balances 51/51 with a brace that does real work rather than a cosmetic pair — a truncated payload whose prefix matches now fails too, where before it verified. Safe for any future additive field: a JSON object's own closing brace is always the last character, whatever type the last value has, which is pinned by two negative-space tests (a nested object and an array-valued last key must both still verify). Cost: +3 bytes, against the 1,873 the resolver fold already gave back. The alternative considered and rejected was dropping the literal `{` for a `?` glob. It balances too, but weakens the anchor from "must be an opening brace" to "must be any one character", and the anchor is the entire point. Two guards added so this cannot recur silently: - runtime-launcher-parity (F0) pins brace balance at the SNIPPET, so the next edit to that pattern fails on the file it broke instead of surfacing three files downstream in a test whose name mentions neither the launcher nor this issue. It also asserts depth never goes negative, since a `}` preceding its `{` nets to zero while being unbalanced at every prefix. - runtime-identity gains behavioral truncated-payload and trailing-garbage fixtures, so the added `}` is proven load-bearing rather than merely present. Verified: snippet 51/51 braces; new-project combined net depth 0; the seven other preamble-bearing files with nonzero depth are unchanged from merged next (their own prose, not the preamble, and not in any guard's scan set); all 112 inlined copies and the resolver reference re-synced byte-equal; sync:launcher idempotent on the second run. Refs #3841 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#3841): backfill changeset PR number Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-25 01:05:53 -04:00
Tom Boucher	de95c03f72	fix(#3699): report why a derived frontmatter key was not written, and repair a missing body source (#3846) ... * test(#3699): failing-first coverage for derived-key reporting and the case-D fallback * fix(#3699): report why a derived frontmatter key was not written, and repair a missing body source * fix(#3699): scope session-field writes to ## Session so an archived line cannot absorb the update * fix(#3699): resolve the session writer from body labels only, so a frontmatter key never writes the body * chore(#3699): backfill changeset PR number --------- Co-authored-by: sim <sim@local>	2026-08-24 23:20:42 -04:00
Tom Boucher	36375513b9	feat(#3840): generate docs/FEATURES.md from per-feature fragments (#3845) ... * feat(#3840): generate docs/FEATURES.md from per-feature fragments docs/FEATURES.md was hand-maintained, and every feature PR wrote into two shared mutable cells: the '### N.' heading whose integer was hand-allocated at authoring time, and the hand-maintained table of contents. Concurrent PRs all picked the same next integer, and two PRs adding differently numbered features still collided on the TOC. #3831 was renumbered 165 -> 166 -> 167 -> 168 across successive rebases, each collision also costing a full matrix verification run because the sha-keyed pass marker dies with the rebase. Mechanism: one fragment per feature at docs/features/<slug>.md carrying id/title/group (and an optional order) in frontmatter, consolidated by scripts/gen-features.cjs --write|--check into a marker-delimited region of docs/FEATURES.md that holds BOTH the TOC and every section body. Group headings and their order are derived too - a group sorts by its lowest-ordered member - so there is no shared registry to edit either; optional per-group prose lives in docs/features/_groups/<slug>.md. A contributor adds exactly one new file. Wired into regen:derived and lint:generated-sync alongside the eight existing generators, matching gen-adr-index.cjs's CLI shape and typed-REASON reporting. Migration froze all 168 existing numbers verbatim: identical section set, identical order, identical bodies. Two defects found in the tree are fixed inline rather than carried forward - the '## Related' block had been spliced into the middle of the document, orphaning §142's Reference line, and four inbound anchors were already broken on next (FEATURES.md#runtime-identity in two files, and #143-spec-phase-edge-completeness-probe off by one). Since the repo has no link checker, --check now validates every inbound FEATURES.md#anchor by resolved target, so that class cannot ship silently again; locale FEATURES.md files resolve elsewhere and stay out of scope. Refs #3840 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3840): carry upstream §69 delta into its fragment and harden the generator Review found section 69 missing '[--strict]' and REQ-STATE-05/06 versus origin/next. Root cause was a stale base, not extraction loss: those lines landed in 394bf384b (#3844) AFTER this branch forked at 63abcface, and 'git diff 63abcface origin/next -- docs/FEATURES.md' is exactly that hunk. Merging origin/next auto-applied the hunk into the GENERATED region, which --check immediately reported as stale; the delta is now carried in docs/features/statemd-consistency-gates.md and regenerated from there. --write is now fail-closed. It previously rendered the region even with violations outstanding, warning only on stderr and exiting 0, so a '--write && git commit' chain could commit a FEATURES.md carrying two colliding sections. It now refuses and exits 1; --force is the explicit override and says so in the report. The test that pinned the old behavior now pins the refusal, plus the --force override and its scoping. Marker forgery is rejected at two layers. A fragment body containing '<!-- FEATURES:START' or '<!-- FEATURES:END' is a typed body_forges_region_marker violation (fragments and group notes alike), and spliceIntoFeatures anchors the end boundary with lastIndexOf instead of indexOf, so a marker that reaches the document by any other route can only make the generated region grow, never shrink. Matching is on marker PREFIXES, so a decorated variant comment cannot slip past. Symlinked corpus entries are refused with a typed dirent_not_regular_file rather than read. A fork PR could otherwise commit docs/features/evil.md as a symlink to any readable path and have the generator inline those bytes into the committed docs/FEATURES.md on the next regen. Equivalence re-verified with a method that cannot cancel out. The first check extracted both operands with the same body-normalising helper, so anything that helper dropped was dropped on both sides. The replacement runs two independent passes: a global content-line multiset diff with no per-section logic at all (0 gained, 19 lost, all 19 the stale hand-written mini-TOC links this change deliberately deletes), and a per-section byte-exact body diff carrying a coverage assertion that fails loudly per file when the extractor accounts for fewer lines than the file contains. That assertion caught two blind spots in the checker itself. 168/168 sections present, order identical, one intended body difference (§142 regains the Reference line orphaned by the misplaced '## Related' block). Refs #3840 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#3840): backfill changeset PR number Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-24 22:49:00 -04:00
Tom Boucher	394bf384be	fix(#3696): report the last_activity invariant and make the verdict gateable with --strict (#3844) ... * test(#3696): failing-first coverage for the last_activity invariant and --strict exit status * fix(#3696): report the last_activity invariant and make the verdict gateable with --strict * fix(#3696): agree with the real reader on last_activity, and stop reporting structure as truncation * chore(#3696): backfill changeset PR number --------- Co-authored-by: sim <sim@local>	2026-08-24 21:33:48 -04:00
Tom Boucher	6905726d9c	chore(#3833): gate every PR compute lane behind a mergeability preflight (#3843) ... * test(#3833): failing-first suite for the PR mergeability preflight * chore(#3833): gate every PR compute lane behind a mergeability preflight * fix(#3833): assert the preflight gate on parsed yaml and guard status-function if * fix(#3833): run stub-backed cli tests in-process to avoid a spawnsync deadlock * fix(#3833): fail the preflight open when its script is absent at the base sha --------- Co-authored-by: sim <sim@local>	2026-08-24 21:13:55 -04:00
Tom Boucher	63abcface9	feat(#3146): resolve gsd_run so workflows cannot reach a foreign gsd-tools (#3831) ... * feat(#3146): resolve gsd_run so workflows cannot reach a foreign gsd-tools The predecessor package get-shit-done-cc publishes a colliding gsd-tools bin whose phases.clear DELETES where this package's ARCHIVES, and both print success-shaped output against a gitignored .planning/ -- which is how #3129 cost a user 43 phase directories with no error and nothing recoverable from git. The launcher's PATH branch now resolves gsd_run, published only by this package and self-locating via its own symlink chain to the sibling shim, instead of the colliding gsd-tools. A foreign handler becomes unreachable from PATH, and when no gsd_run is reachable the resolver fails closed rather than falling back -- that fallback was the vulnerability. This is smaller than the branch it replaces, which matters: the preamble is inlined into 113 shipped files and agents/gsd-verifier.md sits 2 bytes under a red-line size cap. unset -f gsd_run leads the preamble so a re-source is idempotent. Without it, command -v finds the shell function, returns a bare name, and the resolver falls through to an exit 1 that kills a sourced caller's shell. Adds gsd-tools runtime-identity, a manual diagnostic reporting this runtime's package coordinates over the baked package-identity (#498) and readHostVersion, with a strict total classifier: only a JSON object with an exact packageName verifies, since JSON.parse admits 0/"str"/[]/null/true. An inlined identity assertion was built and reviewed first, then withdrawn -- it breaks five frozen size ceilings and no assertion fits in 2 bytes. Closes #3146 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3146): stop sync:launcher relocating a deliberate preamble placement Pre-existing defect, surfaced by this PR because sync is a no-op unless the snippet content actually changes. transformFile inserts the preamble into the first block that CALLS gsd_run, but gsd-core/workflows/explore.md deliberately places it in a bootstrap-only block that DEFINES gsd_run without calling it -- its own comment explains why: declining the research offer must not leave Step 5's commit call unbootstrapped. Stripping empties that block of calls, so the preamble migrated forward and broke the define-before-use invariant tests/explore-command.test.cjs pins. Reproduced on a pristine origin/next checkout with the base snippet and base file, so this was not introduced here. The insertion target now honours a block that already carried the preamble, falling back to the first calling block for files that have none yet. Adds a behavioral regression test over a two-block fixture. Also updates three runtime-launcher-parity tests that pinned the removed PATH fallback to gsd-tools. Their intent is preserved -- the PATH stub is renamed gsd_run so it is reachable by the new resolver, and the RUNTIME_DIR-wins test still asserts the stub is never invoked. Fixture shebangs move to an absolute /bin/sh, because the fixture PATH is deliberately restricted and #!/usr/bin/env sh could not resolve. Refs #3146 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#3146): backfill changeset PR number Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(#3146): document the FEATURES.md section-numbering practice The monotonically increasing section number in docs/FEATURES.md is the most frequent merge-conflict source in this repo, and it has TWO conflict cells, not one: the ### N. heading and the hand-maintained table of contents. Two PRs adding differently numbered features still collide on the TOC, so renumbering alone does not make a branch safe. This branch alone was renumbered 165 -> 166 -> 167 -> 168 across successive rebases. Adds a CONTRIBUTING section stating the practice: allocate the number last, never pre-emptively renumber, take max+1 after a rebase and update the TOC in the same commit, and never renumber someone else's section. Fork contributors are told explicitly they may leave the number to a maintainer at merge rather than chasing the counter. Agents are told to lease the allocation and to include the file in their published touched set. Records the durable fix as planned rather than pretending it exists: FEATURES.md should be generated from per-feature fragments the way CHANGELOG.md is generated from .changeset/, and the way tests/emitted-drift-acks/ works (#2914). Also renumbers this branch's own section to 168, leaving 167 to the PR already in flight. Refs #3146 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-24 20:57:16 -04:00
Tom Boucher	aaf47c5fc2	fix(#3691): let every reviewer lane take a prompt cap, and make the documented global resolve (#3832) ... * test(#3691): failing-first coverage for the reviewer prompt budget No prompt cap can reach any CLI reviewer lane, by any configuration. Two independent defects compound: all nine `transport: spawn` lanes declare `promptBudgetKey: null`, so `budgetFor` returns on its first line; and the documented global `review.max_prompt_tokens` is advertised in the schema manifest but declared nowhere, so the resolver never materializes it and `budgetFor`'s fallback is dead code. Adds to tests/reviewer-config-federation.test.cjs, which already owns the per-reviewer budget config-set/config-get idiom: - a CLI lane inherits the global cap (RED: reports null) - an http lane with the -1 sentinel inherits the global cap (RED: reports null) - the resolved review surface carries max_prompt_tokens at all (RED: absent) - per-lane overrides the global on a CLI lane - the sentinel boundary: -1 inherits, 0 means do-not-trim and must NOT read as unset, 1 is the smallest real budget — the regression budgetFor's own comment warns about - anti-tightening pins that must stay green: an empty config leaves every lane null, the three existing budgeted lanes are unchanged, and config-set still rejects a per-reviewer key naming something that is not a declared lane - a fast-check property over the resolution contract itself, with -1, 0 and non-finite inputs generated explicitly rather than left to chance Every row was reproduced by hand against the real CLI before being written, so the RED/GREEN split is observed rather than predicted. Refs #3691 * fix(#3691): let every reviewer lane take a prompt cap, and make the global resolve No prompt cap could reach any CLI reviewer lane, by any configuration. Two independent defects compounded. The nine spawn-transport lanes — claude, coderabbit, antigravity, cursor, gemini, codex, kimi-code, opencode, qwen — declared `promptBudgetKey: null`, so `budgetFor` returned on its first line and `review-lane plan` reported `promptBudget: null` no matter what was configured. Each now declares `review.max_prompt_tokens_per_reviewer.<slug>` with the same `-1`-is-unset sentinel the three local-server lanes already use. Separately, the central `review.max_prompt_tokens` was listed in the schema manifest's validKeys and documented as a supported setting, but declared nowhere — the resolved surface is built from capability declarations plus the defaults manifest, and neither carried it. `configGet` returned undefined and `budgetFor`'s documented fallback was dead code. It is now declared with a `null` default, exactly as docs/CONFIGURATION.md already specified, so the default behavior is unchanged: nothing configured means nothing trims. Two things the diagnosis had not predicted, found and fixed while implementing: - `REVIEWER_LANES` in src/review-lane-descriptor.cts is a second, hardcoded registration site that `mergeReviewerLanes` prefers over the capability registry on a slug collision. Editing only the capability files left every CLI lane still null. Both sites now agree. - The generated `gsd-core/bin/lib/capability-registry.cjs` was stale and masked the capability edits; regenerated with `npm run gen:capability-registry` rather than hand-edited. docs/CONFIGURATION.md said "Only lanes that declare a budget key accept one — today ollama, lm_studio and llama_cpp". That is false as of this change and is corrected rather than left to rot. The trim-versus-refuse question the issue raises is deliberately not taken up here: the refusal path already exists for the case that matters — a reviewer whose minimum set exceeds its budget is skipped rather than sent a misleading prompt — and trimming above that floor is the documented, shipped design of the feature. Changing it would alter behavior for the three lanes that already work, which is not what the issue asks for. Fixes #3691 * fix(#3691): document the new global and narrow an invariant this change obsoleted The full suite surfaced two consequences of giving every CLI lane a budget key. `review.max_prompt_tokens` entered CONFIG_DEFAULTS without a matching entry in the planning-config reference, which config-field-docs guards. Documented, including the sentinel semantics a reader needs: a per-lane value overrides the global, `-1` means unset and inherits it, and `0` means "do not trim that lane" and is not unset. The #2797 federation guard asserted that "a lane with no model flag and no host owns no config keys". That held only because budget keys existed solely on the three local-server lanes, all of which have hosts. A lane can now legitimately own a config key for a third reason, so qwen tripped it. The assertion is narrowed rather than weakened: such a lane must still own no model key and no host key, and may own at most its own `review.max_prompt_tokens_per_reviewer.<slug>` — never another lane's. That is strictly more specific in the dimensions that still matter. Proven to still bite: hypothetically giving qwen a `review.models.qwen` key fails it with `model/host: review.models.qwen`. The name and comment cite #3691 for why the premise changed, so a reader sees a deliberate narrowing, not erosion. Checked the sibling assertions in that describe block; the other three do not rest on the obsolete premise and are untouched. Refs #3691 * fix(#3685): port the write-flag content-change contract to its three sibling sites #3685 fixed `phase complete`'s `roadmap_updated` / `state_updated`, which reported `fs.existsSync(path)` rather than whether the transaction wrote anything. Three sibling sites carried the identical defect and are ported here. - `cmdPhaseRemove` reported `roadmap_updated: true`, hardcoded. `updateRoadmapAfterPhaseRemoval` now returns whether the content changed and the flag reports it. #2640/#2974 already fixed `state_updated` at this same call site and left this one behind, so the correct shape was adjacent. - `cmdMilestoneComplete` reported `state_updated: fs.existsSync(statePath)` — byte-identical to #3685's bug in a different command. - `cmdMilestoneComplete` reported `milestones_updated: true`, hardcoded, never consulting the MILESTONES.md write. `gsd-core/workflows/remove-phase.md:100` extracts `roadmap_updated` for display and never branches on it, so the flip from always-true to content-based changes no workflow behavior. Verified by reading the step, not assumed. One trap found while implementing: the obvious in-memory `finalContent !== originalStateContent` comparison — copying `cmdPhaseComplete`'s shipped shape verbatim — gives a FALSE POSITIVE for milestone completion. `platformWriteSync` normalizes Markdown at write time, and the milestone-closure transform regenerates `## Current Position` fresh on every call, so its pre-normalize output always differs from the already-normalized file on disk even when the persisted bytes are identical. The comparison is therefore made against the post-write on-disk content. `cmdPhaseComplete`'s own comparisons are left untouched — their repeat-no-op tests pass, so they are not exposed to this artifact. `milestones_updated` has no reachable no-op: the MILESTONES.md write unconditionally appends an entry every call. Only the true direction is pinned, documented inline rather than faked with a passing test. Refs #3685 * fix(#3685): compare write-flag content through the writer's own normalizer An independent reviewer disproved a claim made while porting #3685's contract to its sibling sites: that `cmdPhaseComplete`'s comparisons were not exposed to the Markdown-normalization artifact already diagnosed in `cmdMilestoneComplete`. `platformWriteSync` normalizes on write — CRLF stripped, blank-line runs collapsed, a blank line inserted after a heading, a single trailing newline enforced. Every flag that compares the PRE-normalization in-memory string against the on-disk pre-image can therefore report a change when the persisted bytes are identical. `cmdMilestoneComplete` had been worked around by re-reading the file after the write; the other sites compared raw strings. All of them now go through one exported seam, `contentChangedAfterNormalize(filePath, before, after)`, which normalizes both sides exactly as the writer does. That removes the extra disk read the milestone workaround needed, and makes the sites agree by construction rather than by four independent implementations of one rule — the divergence the repo names as an anti-pattern. Reachability, stated precisely rather than uniformly: the seam is load-bearing at `cmdPhaseComplete`'s `roadmapUpdated`, `requirementsUpdated` and `stateUpdated`, where section-rewrite logic genuinely regenerates content into a different-but-normalization-equivalent shape. At `updateRoadmapAfterPhaseRemoval` it is defense-in-depth: the no-match branch never reassigns `content`, so the raw comparison was already correct there. The first analysis claimed the reverse; this is the corrected finding. Also fixes an unsound test premise the remote suite caught. The byte-identity precondition in `roadmap_updated is false when ROADMAP.md comes out byte-identical` asserted against a hand-authored, un-normalized fixture — so the very first write reformatted it and the file could not come back identical. The fixture is now written already-normalized, so the assertion compares a normalized pre-image against a normalized post-image and still fails if the flag regresses to a hardcoded `true`. Not platform-specific; it reproduces on macOS too, and the earlier local check simply never exercised it. The sibling true-direction and milestone tests were checked for the same premise and do not share it — they assert `notEqual`, or compare two post-write states produced through the same normalizing seam. Refs #3685 * chore(changeset): backfill PR number for #3691 fragment --------- Co-authored-by: sim <sim@local>	2026-08-24 19:39:51 -04:00
Tom Boucher	c933184b97	enhance(#3172): require a stated failing direction for every automated acceptance command (#3825) ... * test(#3172): failing-first suite for the stated failing-direction probe Pins the <fails_when> pairing walk, placeholder denylist, MISSING sentinel exemption, degraded-read contract, CLI arm and the plan-authoring contract text. RED by construction: the module exports it requires do not exist yet. Executed on the remote runner. * feat(#3172): require a stated failing direction for every automated acceptance command Every runnable <automated> command now carries a <fails_when> sibling naming what output constitutes failure. A command with no expressible failure mode is not an acceptance test: it reads as rigour and is not falsifiable. - verify-command-grounding gains a failing-direction probe sharing the existing <automated> grammar, MISSING sentinel and walk guard rather than copying them - gsd-tools check verify-failure-directions <N> backs it; plan-phase dispatches it and hands the JSON to gsd-plan-checker check 8f - Dimension 8 detail extracted to references to stay under the agent size cap Verified on the remote runner. * fix(#3172): close four review findings in the failing-direction probe - MISSING_SENTINEL_RE matched an env-var assignment prefix (MISSING=1 cmd), so a real command was exempted from the new blocking gate. Tightened the SHARED constant rather than adding a second copy. - Both token regexes scanned to EOF on unclosed openers (O(n^2), 1562ms at 40k). Bodies are now non-crossing; 1ms, byte-identical on well-formed input. The pre-existing AUTOMATED_BLOCK_RE carried the same defect and is fixed here too. - probePhaseFailingDirections reported status 'ok' when one plan was unreadable, conflating 'could not look' with 'nothing to report'. - Extracted the phase-resolution block both check arms had copied verbatim. Also corrects a docs/AGENTS.md dimension list stale since #2401. Verified on the remote runner. * fix(#3172): project the planner rule onto the spawn contract, settle emitted bookkeeping The remote runner refuted the planner-side edit. agents/gsd-planner.md is frozen under a 49152-LF-char cap asserted by four suites and sat at 49,146 — six chars of headroom — so the +537 of authoring rule blew it. #3297/#3645 already settled where such a rule goes: the planner spawn contract in plan-phase.md, beside <tracked_source_paths>. The agent file is reverted to origin/next verbatim. - plan-phase.md gains <failing_direction_contract>; tests row 30 now asserts the contract there and row 30b guards the freeze in both directions - plan-phase.md growth acknowledged by APPENDING to the 3409 fragment, per the precedent that two ack sources may never name the same path - install-tree fixtures regenerated for the three new reference files Verified on the remote runner. * chore(#3172): backfill PR number into the changeset fragment pr:0 -> pr:3825 now that the PR exists. --------- Co-authored-by: sim <sim@local>	2026-08-24 19:05:11 -04:00
Tom Boucher	596540f864	feat(#3227): publish machine-readable state contract at step boundaries (#3824) ... * feat(#3227): publish machine-readable state contract at step boundaries Adds src/state-contract.cts, a best-effort publisher that writes .planning/state.json (contract 1.0.0) at 11 step-boundary commands, so external tools read a versioned contract instead of parsing STATE.md and ROADMAP.md heuristically. Composes existing owners rather than re-deriving: phase rows come from a new locateProgressTable extracted from deriveProgressFromRoadmap (so the snapshot can never disagree with GSD's own progress counters), milestone identity from getMilestoneInfo, and next from classifyProject. Owners are required lazily to avoid the state -> state-contract -> smart-entry -> state require cycle. Also fixes a pre-existing defect in scripts/lint-test-file-count.cjs (maintainer-approved as a second concern): testEffectivePrefix never stripped the suite qualifier, so 65 dotted test files counted against no module and 9 mis-bucketed into a shorter one. Allowlist re-baselined for the 74 files the gate can now see. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#3227): backfill PR number into the changeset fragment pr:0 -> pr:3824 now that the PR exists. Doc-only. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test(#3227): shape hostile-name fixtures away from the scan corpus The two hostile-input fixtures used a literal phrase from scripts/prompt-injection-scan.sh's corpus, so CI's Security Scan redded on this file. These tests assert that an arbitrary phase name round-trips into state.json as inert data -- the property holds for any string, so the injection flavor is illustrative, not load-bearing. Reshaped to a hyphenated fake instruction tag, which stays hostile-looking while matching none of the scanner's patterns. Allowlisting the file was rejected: that mechanism is for suites whose subject IS injection defense, and it would blind the scanner to this whole file permanently. See DEFECT.PROMPT-INJECTION-SCAN-COLLISION. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#3227): ratchet the state-contract mutation floor to its measured score The module was registered at minScore 50, the ratchet's minimum permitted floor for a newly-registered module whose score had not been measured. This PR's own Stryker shard measured 66.25% (run 32769289750, job 97565813640), so the floor moves to floor(measured) - 1 = 65, per the rule the registry documents. 66.25 is below TARGET_MUTATION_SCORE (80), so this stays a ratchet candidate: raise as the tests improve, never lower. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-24 17:56:02 -04:00
Tom Boucher	a2387a0545	feat(#3034): add opt-in parallel reviewer lanes (#3822) ... * test(#3034): failing-first coverage for opt-in parallel reviewer lanes Executes the real invoke_reviewers dispatch block from review.md against a stubbed gsd_run seam rather than pattern-matching the workflow text, so the two properties that actually carry risk are observable: that every lane is joined before aggregation, and that concurrent lanes cannot tear a line in gsd-review-lane-results.jsonl. Concurrency is proven by a barrier fixture, not by elapsed time -- each stub lane blocks until all lanes have checked in, which can only complete if they overlap. Red against the current sequential dispatch, by design. Refs #3034 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * feat(#3034): add opt-in parallel reviewer lanes Reviewer lanes within one review pass inspect the same immutable plan snapshot and have no dependency on one another, but were dispatched strictly one at a time, so a multi-reviewer pass cost roughly the sum of its lanes. The serialization is a deliberate protection against provider rate limits, so it stays the default; review.parallel_lanes opts a project out of it. The loop body is hoisted into run_review_lane so the sequential and concurrent paths share one body -- two hand-synced dispatch bodies is the divergence class ADR-2782 spent a phase deleting. Each lane writes a slug-scoped result file, concatenated in selection order after the join: concurrent O_APPEND is atomic only below PIPE_BUF, and write_reviews parses that JSONL to render the models:/model_sources: frontmatter, so a torn line is a broken REVIEWS.md rather than a cosmetic log defect. Aggregating in selection order also keeps the artifact byte-identical between the two paths. The guard is strict equality on "true" and falls back to sequential when config-get fails -- the opposite polarity from the commit_docs guard, because failing open here fires the very requests the default prevents. Also corrects docs/COMMANDS.md and its four locale mirrors, which described --all as running every configured reviewer in parallel when dispatch was in fact sequential. Closes #3034 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3034): de-duplicate dispatch slugs and scope lane locals Review finding (Standards axis): a slug repeated in SELECTED_REVIEWERS would put two concurrent background jobs on the same > -truncated per-lane result file. The shared-append form this replaced could not corrupt itself that way, so de-duplicating is what keeps the concurrent path no worse than the sequential one. Selection de-dupes today -- the roster is a Set and review.default_reviewers normalizes lowercase-unique -- but reachability analysis is not a contract, which is the same reason the roster derivation itself is guarded. Splitting once into DISPATCH_SLUGS also removes the duplicated tr-split the same review flagged: the dispatch and aggregation loops now share one list, which is what guarantees they walk the same slugs in the same order. A plain string accumulator rather than an array, because zsh and bash disagree on array indexing and this block runs under both. Also scopes run_review_lane's locals. Not a live fix -- each dispatched call already forks its own subshell -- but it makes the isolation a property of the function rather than of the dispatch mechanism happening to fork. Refs #3034 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test(#3034): acknowledge review.md growth, drop spent 2295 ack The differential attribution gate reported review.md growing 4173 bytes (30712 -> 34885) with no live acknowledgment. Adds the per-PR fragment it asks for, naming only the one path it reported. Deleting tests/emitted-drift-acks/2295-resolved-model.json is required, not opportunistic. That fragment declared review.md and nothing else, and its ripple is already absorbed into the base, so it is spent -- it can no longer clear anything, which is why the gate still reported review.md as unacknowledged. It could not simply be left alone either: two ack sources may never name the same path, so it blocked this PR's fragment outright. CONTRIBUTING is explicit that a fragment whose last entry is removed gets deleted with it, because an empty fragment signals nothing while its presence reads as a live alarm. Refs #3034 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#3034): backfill changeset PR number Refs #3034 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-24 15:25:22 -04:00
Tom Boucher	a84f756303	fix(#3078): sweep all-spent ack fragments on next, name the collision remedy (#3823) ... * fix(#3078): sweep all-spent ack fragments on next, name the collision remedy `guard-no-ack-on-next` only ever watched the legacy tests/emitted-drift-ack.json. #2914 exempted the fragment directory on the premise that a persisting fragment "cannot conflict with any other PR". Fragments do not share a FILE, but they do share a PATH KEY SPACE, and a path claimed by two sources is a hard failure in the same script -- so a fully-spent fragment on next owns keys it can no longer gate, and the next PR to grow one of those paths can declare it neither there (spent) nor in its own fragment (duplicate). Measured at the sweep: 45 fragments owning 403 paths, up from 13/272 at triage 19 days earlier. - `assertNoAllSpentFragments` fails a fragment only when EVERY surviving entry is spent against the copy at HEAD^, so a partially spent fragment -- and the re-arm-by-appending route #2639/#2993 ship on -- keeps working. - `ackProse` duplicates the gate's zero-width/whitespace stripping across the scripts-ship/tests-do-not line, bounded by a prose-parity test. - The guard job's checkout takes fetch-depth: 2; at depth 1 HEAD^ is absent and every fragment reads as brand-new, i.e. the guard passes vacuously. - The duplicate-ack error now names both resolutions, since the guard is post-merge by design and cannot stop the colliding PR. - All 45 spent fragments deleted, 0000-legacy-migration.json included, and the three tests that pinned its permanence corrected. Verification is the remote runner (gsd-test), not a local suite. Closes #3078 * fix(#3078): make the prose-parity test two-sided, cover the git seam, base on the pre-push tip Three review findings, all fixed: - The parity test was a tautology: it checked ACK_INVISIBLE against a hardcoded list matching its own definition, never against the gate. The gate's INVISIBLE and its reason normalizer (hoisted out of diffEmitted as normalizeAckReason) are now exported for that sole purpose, and the test sweeps 0x00-0xFFFF against both surfaces. Mutation-checked: adding a codepoint to one side and not the other now fails. - resolveBaseRef, readFragmentAtRef and assertUsableBaseRef had zero direct coverage -- the tests reimplemented the git reads in a local helper, so the ls-tree-vs-show discrimination, the root-commit fallback and the option-injection guard were never executed. All are exported and tested against real temp repositories now, plus an end-to-end --base-ref subprocess. - HEAD^ is not 'the state of next before this push'. The default branch allows REBASE merges, so one push can carry N commits, and a 2-commit rebase-merge whose first commit adds a fragment would be told to git rm it on the very push that introduced it. CI now passes github.event.before via --base-ref and fetches it explicitly; HEAD^ remains only the local fallback. Also adds the safe.directory guard every other git call in this repo carries (#2767), and stops naming the deleted migration fragment by filename in CONTEXT.md, which tripped lint-removed-but-needed. Refs #3078 * fix(#3078): keep the fragment directory alive after the sweep empties it Sweeping every fragment leaves the directory untracked, and check-glossary-refs then fails: CONTEXT.md references tests/emitted-drift-acks, which no longer exists. The empty directory IS the intended steady state, so it has to survive its own remedy. Adds tests/emitted-drift-acks/README.md documenting the create/use/delete lifecycle where a contributor actually meets it, matching the existing tests/qa/smell-acks/README.md precedent. Every reader filters on .json, so the README is invisible to the gate. Also sweeps #3809's ack fragment, which the rebase onto origin/next brought in and the new guard immediately reported as all-spent -- its own remedy applied. Refs #3078 * fix(#3078): guard the added tests' git calls, drop a second fragment-existence pin Both defects surfaced by the remote runner (linux-node24, 4/37445 failed). - The new --base-ref E2E test ran `git rev-parse HEAD` against the checkout without the #2767 safe.directory guard. The runner mounts the repo at a path owned by another uid, so git refused every operation there with 'detected dubious ownership'. Every git call the new tests make now names its own specific directory as safe, via one local helper, mirroring safeDirArgs in helpers/emitted-runtime.cjs. - tests/agent-tracked-source-rule.test.cjs pinned the existence and contents of the 3645 and 3409 ack fragments. That is a merged PR's paperwork, not live behavior: once the growth is in next's baseline the acks are spent and this PR's guard sweeps them. The third assertion pinned the hand-appended workaround for the exact collision #3078 removes. Deleted; #3645's real protection is the two behavioral tests above it, untouched. Also restores #3809's ack fragment, which merged one commit before this branch. Deleting an ack in the same window as its introducing PR races any consumer whose baseline predates it -- the runner's container proved it, resolving origin/next to 8ed105c8a where the file is still 13847. The backlog sweep is this PR's scope; that fragment is left for the guard's own first run. Adds the rule to the fragment README so the class stops recurring. Refs #3078 * test(#3078): derive the E2E guard expectation from the fragment inventory The --base-ref E2E test asserted exit 0 while passing the checkout's own HEAD as the base ref. HEAD-as-base makes every present fragment byte-identical to itself, so all of them are trivially all-spent and the guard correctly exits 1. The test only ever passed because the directory happened to be empty when it was written; restoring #3809's fragment made it fail. The script was right and the test was wrong. The degenerate base ref is kept deliberately -- it is what makes 'spent' trivially true and therefore deterministic -- but the expectation is now derived from listFragmentFiles() at runtime: zero fragments means exit 0 and the no-survivors line, N fragments means exit 1 with every name and its git rm. Proven state-independent by running the suite with the fragment present, with the directory emptied, and with it restored. The option-shaped --base-ref rejection is split into its own test, unchanged. Refs #3078 * chore(#3078): backfill PR number into the changeset fragment (pr:0 -> pr:3823) --------- Co-authored-by: sim <sim@local>	2026-08-24 15:24:30 -04:00
Tom Boucher	4b84be1da4	fix(#3683): wire gated learnings extraction into completion, align copy path (#3810) ... * test(#3683): failing-first rows for learnings source resolution and wiring pins * fix(#3683): wire gated learnings extraction into completion, align copy path * test(#3683): register the learnings suite in the docs-guard lane, drop unverified markers * fix(#3683): close review findings — per-item parsing, readdir guards, docs paths * fix(#3683): route phase enumeration through the locator seam, fix assertion targets * fix(#3683): merge execute-phase ack into the 3003 fragment, fix fidelity targets * fix(#3663): replace the spent execute-phase ack entry with the 3683 re-arm * chore(#3683): backfill changeset pr number --------- Co-authored-by: sim <sim@local>	2026-08-24 09:23:01 -04:00
Tom Boucher	4af59f8dd3	fix(#3662): resolve managed hook node runners at hook-fire time (#3790) ... * test(#3662): failing-first suite for runtime-resolving hook runners * fix(#3662): resolve managed hook node runners at hook-fire time * fix(#3662): close review findings and document the resolver * fix(#3662): close adversarial and security review findings * chore(#3662): backfill changeset pr number * test(#3662): honor win32 skip return and platform-aware sh runner pin * test(#3662): pin the bare win32-claude sh-hook shape omitting the bash runner --------- Co-authored-by: sim <sim@local>	2026-08-24 00:07:06 -04:00
Tom Boucher	cf15682d1c	enhance(#3028): responsive Markdown separators instead of fixed-width rules (#3789) ... * feat(#3028): responsive Markdown separators instead of fixed-width rules Stage banners, checkpoints, completion and error panels used fixed-width runs of box-drawing characters -- a 53-column heavy rule and a 62-column double-line box. Those runs are ordinary text to a Markdown-rendering host, so in a narrower pane they wrap and the border comes apart from the heading it framed. Shipped content now emits an ATX heading for a titled section and a blank-line-delimited --- for a break between sections, both of which adapt to the available width. The same convention is applied to the three code sites that built these strings at runtime: the UAT checkpoint renderer, the milestone-close audit report, and the TDD review checkpoint table. Removing the box also removes its only reason to exist -- the east-asian-width padding helpers that kept its right border aligned (checkpointBoxLine, displayWidth, isWideCodePoint, ZERO_WIDTH_MARK_RE, CHECKPOINT_BOX_WIDTH). RTL directional isolation is unchanged. The convention is specified in gsd-core/references/ui-brand.md and enforced across all shipped content by tests/responsive-separators.test.cjs. Refs #3028 * test(#3028): pin the heading form in checkpoint and audit-report assertions These suites asserted the exact box borders and the 62-column padded banner interior. With the box gone they assert the ### heading form, the --- break and the bolded instruction line, and each now carries a positive assertion that no box character remains -- which is what pins the fix rather than merely tolerating it. Language coverage is converted, not dropped: Japanese, Chinese, Korean, Hindi and Arabic all still assert their rendered banner, and the Arabic case still asserts the RTL directional isolates the box removal must not disturb. Adds a case for a banner longer than the old inner width, which previously produced a ragged border and now has none. Refs #3028 * chore(#3028): acknowledge execute-plan.md growth from the checkpoint display spec The checkpoint_protocol display spec described the drawn box; it now describes the heading, the --- break and the bolded action prompt, which costs 22 bytes (40111 -> 40133, 827 under the cap). Appended to the existing #3370 fragment rather than filed as a new one: a growth ack keys on the bare filename and #3370 already declares execute-plan.md, so a second source naming it would be a hard duplicate-key error. Same supersede-by-append route #3370 took for the spent #2652 fragment. Refs #3028 * docs(#3028): state the load-bearing half of the separator rule, and amend the zh-CN reference Review found three things. The rule as first written demanded a blank line above AND below every ---. Only the one above is load-bearing: it is what stops CommonMark reading the rule as a setext underline for the line above. The one below is cosmetic, because a thematic break is a leaf block. The rule now says that, with the reason, instead of asserting a stricter form the content does not keep. The zh-CN reference had received the mechanical box-to-heading swap but none of the prose behind it: it still claimed a 62-character checkpoint width and still listed --- among forbidden mixed banner styles, so it contradicted the convention it was translating. It now carries the separator section, the setext reasoning, the unconditional-vs-per-runtime rationale and a corrected anti-pattern list, in Chinese. The user guide asserted that a heading is not a degradation anywhere. That is an assertion, not a demonstration. It now says what was actually traded away in a plain terminal, points at the recorded rationale, and invites the report that would justify the capability flag instead. Refs #3028 * chore(#3028): backfill changeset PR number Refs #3028 --------- Co-authored-by: sim <sim@local>	2026-08-23 22:38:12 -04:00
Tom Boucher	107eb8c1d9	feat(#3753): run docs guards on the PR that changes the docs they read (#3787) ... A PR whose diff is entirely under docs/ runs zero tests, so a guard whose INPUT is shipped prose cannot protect the PR lane of the diffs it exists to check. Its only firing opportunity is after merge, on the shared branch -- which is how next went red on dacae9273 while the PR that caused it (#3746) was green on every check. The docs-lint job in .github/workflows/docs-required.yml -- an ALREADY-REQUIRED context -- now selects and runs the docs guards that read the specific docs files the PR changed. scripts/docs-guard-registry.cjs test file -> the docs paths it reads (63) scripts/select-docs-guards.cjs pure (changedPaths, registry) -> test files scripts/lint-docs-guard-registration.cjs drift guard, wired into lint:ci scripts/ci-test-scope.cjs is NOT touched -- `git diff origin/next --` on it is empty -- so #764's saving stands and its 21 pinning tests are untouched. Selection: exact path; trailing-slash directory prefix (boundary-checked -- docs/adrenaline.md does NOT match docs/adr/, which a naive startsWith gets wrong); and '*' for the 6 entries that walk docs/ generally or read a computed path. Unknown maps to '*' -- guessing narrow is how a guard silently stops running. Measured: a typo fix selects 6 of 63; docs/AGENTS.md selects 12; docs/COMMANDS.md selects 18. Four things this got wrong first, each found by an independent reviewer or by probe, and each having been asserted safe in a comment: 1. The registry started as a RULE in ci-test-scope.cjs's RULES, on the theory that classify()'s !codeChanged normalization made it inert. True for docs-ONLY diffs; false for MIXED docs+code diffs, where codeChanged is true and the normalization never runs: node scripts/ci-test-scope.cjs --files "docs/a.md src/semver.cts" with the RULE: 25 targeted_tests origin/next: 3 targeted_tests Category error: RULES is the scoped lane's input; a docs-guard registry is a lane manifest for a consumer that never calls classify(). Extracted; pinned by value. 2. The second attempt was a dedicated workflow with paths: [docs/**]. Such a workflow never reports on a non-docs PR, so it can never be a required context without hanging every non-docs PR -- and a non-required check does not block a merge, so the guard would have been advisory and #3753 unfixed. docs-required.yml already has no paths: filter, already supplies the required docs-lint context, already computes docs_changed, and already ran one docs guard gated on it. Generalizing that step needs no ruleset edit at all. 3. The registry and the drift lint were built from ONE path-segment heuristic, so both were blind identically -- and blind at the guard that motivated the issue. The reader-call regex required a character BEFORE its keyword, so a callee named exactly read( / load( / parse( / doc( / file( / content( could never match; and only an INLINE path.join(ROOT,'docs','X.md') argument was caught, missing the two-step-via-variable form -- the MAJORITY spelling -- plus template literals and concatenation. Detector 1 fired on 14 of ~450 files, so 35 genuine guards sat unregistered while the lint reported 0 violations, including cursor-reviewer (reads docs/COMMANDS.md, asserts .includes('--cursor')) and inventory-headings-countfree. The "accepted blind spot" this shipped with was the common case, not a fringe. 4. With detection fixed the true population is 115 files: 63 genuine guards, 52 incidental. Running all 63 in a REQUIRED check on a one-line typo fix is the cost #764 exists to avoid -- install.test.cjs is 7840 lines and reads exactly one docs file, docs/AGENTS.md, for its frontmatter. Dropping it reproduces the bug; running it for a typo elsewhere is waste. Hence the map. Then a second review round found six more, all fixed here: - fragment-single-edit-propagation.install.test.cjs was EXEMPTED as "overlay fixture only". False: it reads the real docs/registries/eos.json and asserts on a registry entry name, and reads the real ADR-0001 and asserts its H1. A docs-only PR touching either would have gone green and red next -- #3753 shipping again, from inside the fix for it. Now registered against both paths, and all 52 remaining exemptions were re-audited one by one. - The SUITES-collision guard compared RAW registry keys, but run-tests.cjs strips a leading `tests/` BEFORE its suite check. So it caught 'all' and missed 'tests/all' -- the only spelling that can actually occur, since every key carries the prefix. One typo would have run all 824 test files inside the required job. Now normalized the same way run-tests.cjs normalizes. - The lint failed OPEN on an unreadable tests dir or candidate file: 0 violations, ok:true. A guard that cannot read its input must never report success. - The exemption ratchet gated identity only, so a baselined file that later STARTED asserting on shipped docs stayed exempt silently -- 52 permanently blind files. The baseline now fingerprints the docs paths each exempted file references and fails when that set changes, naming what changed. - The exemption marker was still honored inside a multi-line template literal in the header window. The scanner now tracks template-literal and block-comment state. - `git diff --name-only | grep '^docs/'` silently dropped C-quoted non-ASCII docs paths, making docs_changed=false a green zero-guard check. Both call sites now pass -c core.quotepath=false. - The run step was gated on hashFiles(), which a force-committed .docs-guard-tests.txt would satisfy. The step now rm -f's both scratch files first and gates on an output it sets itself. Three empty states, deliberately distinct, because conflating them rebuilds #3753: an empty or malformed registry HARD-FAILS; docs changed with no guard covering them logs and skips; no docs change is already gated. The middle state must never be expressed as an empty --files-from, which prints `no tests in suite "all"` and exits 0 -- a green check that guarded nothing. With the current registry that state is unreachable, because the six '*' entries always match; the branch is kept as defensive handling for a future registry and says so. timeout-minutes: 15 bounds the required job against a hanging fork-supplied test; it had none. npm ci was added because the job never installed dependencies -- the previous single-file step got away without it, the registry does not. docs/contributing/docs-guard-registration.md documents the rule, following its sibling cross-platform-portability-rules.md, and CONTRIBUTING.md's CI Test Quality Checks table links to it. It is also load-bearing: without a docs/ file in the diff this PR would not have triggered its own lane, shipping an unexercised change to a required check. One unrelated fix, included because this PR surfaced it and CLAUDE.md forbids deferring a defect found while working. On this branch's first CI run, `full test (windows-latest, 24, shard 3/3)` was CANCELLED at exactly 30 minutes; tests were still passing 0.8s before the cancel, so it is a wall-clock timeout, not a hang, and a cancelled job reddens `Required tests`. The cause is not this PR's test file, which costs ~60ms. Shard composition is unstable: adding ONE file to the unit suite reshuffled 115 of 268 files between shards, and shard 3 drew a heavier mix. Underneath that is a real pre-existing defect. tests/ci-test-job-timeout-budget.test.cjs requires every lane's budget to be >= 1.5x its MEASURED cost -- "a lane that got slower must be re-budgeted, not excused" -- and its test-full entry recorded 19m from a windows-22 shard. That is stale. Measured on `next` with none of this PR's changes present: 26m18s (run 32614439702, windows-latest/24 shard 3/3), 23m36s and 23m17s on shard 2/3. So the lane costs ~26m and the 30-minute cap carried 1.14x headroom, not 1.5x. The gate had been out of compliance with its own rule; this PR was merely the file addition that reshuffled shard 3 past the cliff. Fixed as that file prescribes: measuredMinutes 19 -> 27 with fresh evidence, and test-full timeout-minutes 30 -> 45. The rule's minimum for 27m is 41; 45 is deliberately above it because the reshuffle means per-shard worst case moves run to run, and a budget pinned to the exact minimum would be re-breached by the next test file anyone adds. Only that one job's timeout changed; test.yml's scope, matrix and steps are untouched, so #764's saving is unaffected. Raising that cap let the Windows shard finish (28m45s, inside 45) and uncovered a real failure the 30-minute cancel had been masking: `new quick-task branch branches off origin/main (#2916)` died with `outcome=timed_out exitCode=null`, SIGTERM, at the 15000ms bound. tests/quick-branching.test.cjs:149 `runStep` runs a `#!/usr/bin/env bash` script executing MULTIPLE git commands, but was bound to GIT_TIMEOUT_MS (15000) -- the norm for a SINGLE git plumbing call. tests/helpers/timeouts.cjs already documents this exact failure and exists to fix it: HOOK_FANOUT_TIMEOUT_MS was created after PR #3285 recorded "outcome=timed_out exitCode=null at exactly the 15000ms probe bound while every other lane passed the same commit", and calls that "a bound sized for the wrong class, not a slow machine". Our failure is that case verbatim, so both sites move to the class norm rather than to a bigger number. The same class also failed on `next` itself 21 hours earlier -- run 32608945654, windows-latest/24 shard 1/3, `plan touching only src/ in a submodule project keeps worktree isolation ENABLED` -- where tests/worktree-safety.test.cjs:5845 `runGate` fans out to `git config --file .gitmodules` under a hardcoded 30000. Fixed too, since it is a defect in the tree regardless of which branch surfaced it. A survey of the whole tests/ tree found the same class-mismatch at further bash fan-out sites bound under 60000ms, and the maintainer approved sweeping them rather than leaving them latent to surface the same way one at a time. 16 fan-out sites across 16 files now use the class norm. The sweep is class-correctness, not raising numbers until things pass. Sites were moved ONLY where the bash body demonstrably spawns something (git, node, npm, a CLI); self-contained shell snippets were left where they are, and are listed as deliberately unchanged: pure if/printf bodies (copilot-install), pure array/case builtins (code-review-pipeline-regression:638), a documented pure-shell gsd_run stub (host-integration), single-process hook calls (workflow-guard:222/271/302), and a deliberately tight 5000ms fast-check hook (gsd-write-guard.property). Nothing was lowered. process-seam.test.cjs:513 (literal 300) is untouched on purpose -- it tests timeout BEHAVIOR, so raising it would destroy what it asserts. Shared file-level constants were the trap here, and were handled per file rather than by redefinition: GIT_TIMEOUT_MS has ~15 users in git-base-branch and only 1 is a fan-out; WORKTREE_TIMEOUT_MS has 16 users in worktree.test.cjs and 3 are; PROBE_TIMEOUT_MS has several in three more files. In each the CALL SITE was changed and the constant left alone, so no single-plumbing-call site silently inherited a 60s bound. The one exception is hooks-opt-in.test.cjs, where HOOK_TIMEOUT_MS has exactly one consumer -- spawnHook, the fan-out itself -- so redefining it is identical in effect and reads better. Only two of these sites have actually been observed failing. The rest cite that shared class and those two run ids rather than inventing evidence of their own. Co-authored-by: sim <sim@local>	2026-08-23 21:21:21 -04:00
Behruz Nassre Esfahani	622f43353c	fix(#3299): tracer feedback gate honors workflow.human_verify_mode (#3390) ... * fix(#3299): tracer feedback gate honors workflow.human_verify_mode The tracer feedback gate (#2294) predates `workflow.human_verify_mode` (#3309, whose scope was the planner and verifier only), and branched on auto-mode alone. Under the documented `end-of-phase` default an interactive run therefore halted after EVERY `type="tracer"` task, synthesizing a `checkpoint:human-verify` no planner ever emitted and asking the user to retype a verdict the executor had just computed — at the cost of a full executor cold-start each time. Planner-side suppression cannot reach this halt because the executor synthesizes it at runtime, which is why #3309 did not close it. The gate now branches on HUMAN_VERIFY_MODE in the interactive path: under `end-of-phase` an automated-only tracer `<verify>` is re-run and, on success, expansion continues with no checkpoint. HALT-on-failure is unchanged. `mid-flight`, `gate="blocking-human"`, and tracers carrying genuine `<human-check>` evidence all still stop; the autonomous branch is untouched. `--default end-of-phase` on the config read is load-bearing, not decorative: `workflow.human_verify_mode` is absent from SCHEMA_DEFAULTS, so a bare `config-get` exits non-zero with `Key not found` on any project whose config.json predates #3309 — which is the reporter's exact config and every pre-existing project. Both copies of the rule (workflows/execute-plan.md and agents/gsd-executor.md) are updated together; the reference doc records the seam and the human-check-still-halts rationale so it cannot recur. Fixes #3299 Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * chore(#3299): add changeset Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(#3299): reconcile the canonical schema table and the stale acceptance test Review round 1 (trek-e) — three items, all in the drift class this PR is about, two of them landed inside this PR's own diff. 1. docs/reference/plan-md.md:233 — CONTEXT.md names this file the canonical schema reference for the tracer task-type contract, and its Task-types row still claimed interactive runs unconditionally present a checkpoint:human-verify. CONTEXT.md and docs/AGENTS.md were updated in the first round; this one was missed, so the authoritative reference was the wrong answer. The row now carries the human_verify_mode-conditional behavior and points at the canonical precedence chain. 2. tests/tracer-bullet.test.cjs — the docs assertion only checked that a tracer ROW EXISTS, never its content, which is why CI could not see the drift. It now asserts the row's actual claims and rejects the pre-#3299 wording. Separately, the #1945 acceptance test named 'interactive run emits checkpoint:human-verify after the tracer' kept passing only because its substrings still occur in the fallback clause, while its name asserted the opposite of shipped behavior. Renamed and narrowed to what #1945 still guarantees, plus a new interactiveIsConditional pin so the unconditional prose cannot be restored under a passing substring check. 3. plan-md.md's <verify> row now documents that the legacy bare-text form (valid, and still shown at :179) does not reach the #3299 auto-continue — only a <verify> carrying <automated> does — so the benefit is silently unreachable for tracers using that format. Mutation-verified: reverting the plan-md row fails 1 test; reverting the executor's interactive branch fails 4. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(#3299): make the tracer gate reachable from the planner template, and bind the assertions Peer review round 3 found two Majors, both verified by reproducing the mutation before fixing. MAJOR 1 — the fix was largely inert on its own default path. agents/gsd-planner.md's Nyquist Rule (:191) says every <verify> includes <automated>, but the tracer-specific template twelve lines later emitted the legacy bare-text form. The gate auto-continues only on a <verify> carrying only <automated>, so every tracer produced from the canonical template fell to the STOP fallback and #3299's benefit was unreachable for exactly the task type it targets. Template now wraps in <automated>; a contract assertion pins it so the two cannot drift apart again. MAJOR 2 — the new assertions did not bind condition to action. Appending 'Nevertheless, interactive runs always present a checkpoint:human-verify' to the canonical row, and 'then immediately STOP and return a checkpoint:human-verify' to the auto-continue clause in BOTH operative copies, restored unconditional interactive checkpointing and left the suite 35/35 green. Every required keyword still matched. Fixed by: - clause 2 must now contain no STOP outcome and emit no checkpoint at all — 'never a checkpoint' has to be true OF the clause, not merely stated in it; - interactiveIsConditional replaced with the ordered-clause parse plus the same no-STOP property, instead of proving only that HUMAN_VERIFY_MODE appears somewhere on the line; - the plan-md.md Autonomy cell is now pinned EXACTLY rather than by keyword presence. Deliberately brittle: CONTEXT.md names that table the canonical schema reference, so a wording change must be a conscious edit in both places. Mutation-verified after the fix: the combined semantic regression now fails 3 tests; reverting the planner template fails 1. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * test(#3299): exact-pin the safety clauses instead of blacklisting outcome verbs Peer review round 4. Blacklisting did not hold, twice over: - Round 3 banned literal STOP and the 'return a'/'present a' checkpoint forms in the auto-continue clause. Round 4 defeated that by appending 'then pause and invoke checkpoint_protocol with a checkpoint:human-verify before expansion' — none of the banned tokens, same restored interruption after every successful tracer. 36/36 passed. - The planner guard looked for <automated> anywhere inside <verify>, so '<verify>[...]<!--<automated>--></verify>' satisfied it while leaving the legacy bare form operative. 107/107 passed across tracer, planner and the three size-cap suites. Synonyms are unbounded; the clauses are not. Both are now pinned exactly on normalized whitespace, the same approach already proven on the plan-md.md Autonomy cell, with defence-in-depth checks behind them: no checkpoint-emitting or blocking outcome in any wording inside clause 2, and the planner's <verify> body must be exactly one non-empty <automated> child with no commented markup. These pins are deliberately brittle. Each is a safety contract, so changing the behavior must be a conscious edit in both the prose and the expectation. Mutation-verified: the synonym-checkpoint mutation fails 1; the commented-out wrapper fails 1; the round-3 literal-STOP + contradictory-doc-row regression fails 3. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * test(#3299): strip comments, require uniqueness, pin whole regions Peer review round 5. Exact-pinning one clause was still bypassable two ways, both reproduced before fixing (each left the suite fully green): - COMMENTED DECOYS. Put the correct text in an HTML comment followed by a live wrong copy: every extractor selected the commented decoy. Worked against the planner template, the canonical plan-md.md row, and both executor branches. - SURROUNDING OVERRIDE. Insert 'after every tracer, pause and invoke checkpoint_protocol before expansion, regardless of the mode-specific rules below' immediately ABOVE the pinned clause, or 'ignore row 3; always wait for approval' below the canonical table. The pinned text was untouched, so equality held while the shipped meaning inverted. The shape that holds, applied to every operative surface: 1. strip HTML comments BEFORE selecting, so a decoy cannot be chosen; 2. require the structural anchor to occur EXACTLY ONCE, so a live second copy cannot hide behind a correct first one; 3. pin the ENTIRE decision region, not one clause, so no unparsed prefix or suffix can override what the pin proves. Applied to: the executor's whole tracer branch, execute-plan.md's whole dispatch line, checkpoints.md's whole precedence section, and plan-md.md's Autonomy cell. Also addresses the round-5 Minor: the planner template is now asserted STRUCTURALLY (exactly one <verify> in the fenced block, body exactly one non-empty <automated> child) rather than pinning the descriptive placeholder verbatim, so behavior-preserving wording changes no longer false-fail. The clause and section pins keep their exact form — those have a safety rationale the placeholder copy does not. Mutation-verified, all six rounds: override-above-clause 1; commented decoy row 1; commented decoy branch 1; ignore-row-3 override 1; synonym checkpoint 1; commented-out wrapper 2. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * test(#3299): drop the superseded exact-placeholder planner assertion Peer review round 6, Minor. The round-5 brittleness fix ADDED a structural planner assertion but left the old exact-placeholder one in place, so the over-brittleness it was meant to remove was still live: rewording the descriptive placeholder while preserving exactly one non-empty direct <automated> child failed the old test and passed the new one. Removed the old test. The structural assertion is the real contract — the gate auto-continues on the SHAPE of the verify, not on the wording of a placeholder. Verified both directions: a behavior-preserving reword now passes; reverting the template to bare <verify> still fails. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * test(#3299): select operative prose via parsePredicates, not a hand-rolled scanner Peer review round 7. I had judged the round-6 selector bypass adversarial-only and out of scope, intending to disclose it. Both premises were wrong, and the review said so: - 'Needs new src API' — false. parsePredicates is ALREADY a public export and internally uses the repo's interleaved fence/comment scanner. Instrumenting candidate lines as throwaway predicate declarations borrows that scanner with no src change at all. - 'Adversarial-only' — false, and this is the part that mattered. Two ORDINARY edits silently turned the guards into decoy checks: * a forgotten '-->' comments the live rule through to EOF, and the balanced-only stripper still saw and accepted the commented rule; * a normal fenced documentation example of the rule, plus a whitespace-only reformat of the live list item, made the selector choose the example. Neither needs intent. A dangling comment is a typo; a fenced example is good documentation. Together they reproduce exactly the accidental drift #3299 came from — with CI green. The selection layer now defers to parsePredicates for operativeness, uses whitespace-tolerant anchors so a reformat cannot decouple the live line from its pin, extracts regions by operative line index rather than string search, and carries a self-guard test proving fenced / balanced-commented / after-unclosed-comment copies are all excluded. The helper also ignores indexes it did not inject, so a pre-existing GSDTEST.CANDIDATE line cannot pollute it. Verified both ordinary-edit scenarios now fail the suite (each was green before). Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * test(#3299): close the operative-selection gaps the maintainer blocked on trek-e's Blocker: the operative-line selection layer had three gaps, all reachable by ordinary future doc edits rather than sabotage. He independently found a fourth I had not disclosed. All are fixed. 1. INDENTATION PROMOTION (his find, not in my disclosure). The instrumentation replaced a matched candidate with an UNINDENTED marker regardless of the original line's indentation. A 4-space-indented CommonMark code block is not skipped by parsePredicates (it accepts indented declarations by design), so stripping the indent PROMOTED an indented decoy to operative — the exact inversion of the guard's purpose. The marker now preserves the original indent, and a candidate that is itself indented 4+ spaces is never injected. 2. NO SET MEMBERSHIP. The filter accepted any in-range integer, so a pre-existing literal GSDTEST.CANDIDATE=<valid index> in source text could pollute the count. Now filters on a Set of the indexes actually injected on this call. 3. RAW FENCE SELECTION (planner). The template test matched the first raw ```xml fence after the marker with no fence/comment awareness — the one selection in the suite that was not operative-aware — so a commented-out decoy template between the marker and the real one would be selected while the live template regressed. The opener must now be operative AND the first non-blank line after the marker. 4. RAW END ANCHOR (regionFrom). The end anchor was tested against raw lines, so a fenced example containing a ### / <type line truncated the pinned region early — a false FAILURE on a legitimate doc edit. End anchors now go through the same operative filter as start anchors. Mutation-verified: the indented-decoy + whitespace-varied-anchor combination and the commented-out fence decoy each now fail the suite (both passed clean before). Truncation is confirmed fixed by extraction — the region spans the full section and retains the content following a fenced example, where it previously stopped at it. Note on the remaining brittleness: adding a fenced example INSIDE a pinned region still fails the whole-region exact pin. That is the intended tradeoff for a safety contract, not the truncation defect, and is called out as such. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * test(#3299): allow-list operative indentation; pin marker provenance Review round 9. BLOCKER — the round-8 indentation guard was written as a DENY-list, /^(?: {4,}|\t)/, and CommonMark has more indented-code forms than that enumerates: " \t", " \t" and " \t" all open an indented code block and all slipped through, so an indented decoy was still promoted to operative while the live rule regressed (34/34 green). Inverted to an allow-list — only 0-3 literal spaces is ordinary block indentation; anything else is code. Enumerating the bad shapes was the error, not the specific regex. MINOR — the injected-index Set validated the marker's VALUE but not its SOURCE. A pre-existing literal `GSDTEST.CANDIDATE=<n>` could name an index that some other (skipped) candidate had contributed to the set, and be accepted. Now also requires p.line - 1 === Number(p.value): the predicate must have been parsed from the line it names. MINOR (false negative) — ```xml title=x is a valid CommonMark info string, and requiring exactly ```xml failed the suite (33/34) on a behavior-preserving edit. Both the opener assertion and the extraction now accept an info string. Mutation-verified: the mixed " \t" decoy and the forged-provenance marker each now fail; the info-string fence no longer false-fails. KNOWN LIMITATION, disclosed on the PR rather than papered over: parsePredicates is a predicate parser, not a general CommonMark operativeness oracle. Two standards-valid constructs still read as operative — a lazy blockquote continuation line (state opens only on a line that literally starts with ">"), and a comment opened mid-line ("prose <!--", where state opens only when the trimmed line STARTS with "<!--"). Closing those means either teaching the shared src/context-predicates.cts about container/lazy-continuation state — a change to a module every health rule consumes, well outside a tracer-gate fix — or hand-rolling a CommonMark parser inside a test, which is how this suite got into trouble in the first place. Left for the maintainer to scope. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * chore(#3299): re-arm the execute-plan.md emitted-drift ack after the base merge The #3299 ack rode on tests/emitted-drift-acks/2652-quick-diagnose-dispatch-isolation.json, which upstream retired in 362d0434b (#3370) once #2728's entries were spent. #3370's own fragment now owns execute-plan.md at the base, so a new 3299-*.json naming that path would collide — mergeAckSources rejects a duplicate key across fragments rather than silently last-winning. Re-arms #3370's entry instead, the mechanism the gate is built for (a spent ack whose reason changes in the diff is live again), carrying #3370's own reason forward verbatim so the base growth keeps its account. Verified: emitted-attribution 175/175 against origin/next@be9329b10. * fix(#3299): honor golden rule 6 in the tracer gate, extract the chain Addresses the review on #3390 (B1-B3, M1-M4, minors). B3 — checkpoints.md asserted two incompatible rules about the same gate. Golden rule 6 says gate="blocking-human" stops for a human in every mode; the precedence table scoped row 1 to interactive runs, so a first-match chain let an auto-mode tracer carrying that gate fall to row 2 and auto-continue. Rule 6 wins: row 1 is now "Any run, any mode", the justification sentence it falsified is gone, and the STOP is evaluated before the auto-mode branch at all three dispatch sites — gsd-executor.md, execute-plan.md and the plan-md.md schema row. Unreachable by our planner is not unreachable: src/verify.cts parses only `type` and never consults `gate` on non-checkpoint tasks, so an imported PLAN.md can carry it. B1 — the LARGE-tier cap. gsd-executor.md is 49150 on next against a 49152 cap, so this PR could not add a byte. Extracted rather than trimmed: the precedence chain now lives only in checkpoints.md (already @-imported by <checkpoint_protocol>, so no new load), and the duplicate summary inside that protocol section is a pointer. The rationale the earlier trim deleted is restored — "production-quality, never a throwaway" and "Pouring more layers onto a broken foundation...". Result 49097: 55 bytes under the cap and a net 53-byte REDUCTION against next, so the PR returns headroom instead of consuming it. B2 — merged upstream/next and resolved all three drift-ack conflicts. 2775 changed shape upstream (string -> {reason}); adopted the new form. M1 — the 2775 ack claimed the Nyquist Rule sat "twelve lines earlier"; it is ~75 lines. Corrected to "earlier in the file". M2 — ack arithmetic restated from measurement, not from a stale base. The 2943 #3299 append is DELETED: with gsd-executor.md now shrinking there is no ripple to acknowledge, and emitted-attribution correctly flagged the entry as stale. M3 — changeset rewritten to the documented bold-lead + em-dash one-liner. M4 — the two self-defeated shapes are gone. The planner-human-verify-mode presence checks now go through operativeLineIndexes. The config-get check does NOT: all three reads live inside ```bash fences, which is their correct executable form, and that selector excludes fenced lines by design. It instead pins exactly one live, uncommented, fenced read per file — mutation-tested against both a commented-out read and a duplicate. Minors — dangling colon lead-in dropped, a "below" pointer that pointed above corrected, and the `(default)` asymmetry between the two dispatch copies aligned. Two defects the merge surfaced, both caught only by the full suite: the new #3576 gate rejected this PR's own bare `references/checkpoints.md` cite in planner-human-verify-mode.md (rewritten to the canonical gsd-core/ form), and the line-keyed PROSE_ALLOWLIST entry for gsd-executor.md needed 794 -> 795 after this change shifted the line. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(#3299): correct the size record the 08-22 merge falsified Review round: one Major, four Minors. Major — the #3299 arm's arithmetic was measured before the merge and is now wrong in a document whose whole purpose is to be an accurate size record. Re-measured at head: execute-plan.md is 39315 B on next and 40111 B here, so the 796-byte delta was right but the endpoints and the headroom were not (849 bytes against DEFAULT_CAP 40960, not 1003). The superseded figures are named rather than silently replaced. Confirmed the workflow cap counts LF BYTES while the agent cap counts CHARACTERS — two caps in two units, one per file. Minor 1 — 2943-context7-tool-name.json reverted to next. JSON.parse of both sides was already identical; the diff was an em-dash/times-sign re-serialization left over from adding and then removing the #3299 arm. No business in this PR. Minor 2 — the duplicated `tracer row Autonomy cell` test is gone. Both copies were new here and carried the same ~8-line canonical string; the one removed selected its row with a raw startsWith find, the shape this suite records at :477 as defeated in round 1. Its rationale — why the cell is pinned EXACTLY, and the append-a-contradiction attack that defeated keyword matching — is carried onto the surviving fence-aware copy rather than deleted with it. Minor 3 — the executor's condensed interactive clause said only "re-run, continue", which does not distinguish pass from fail; read in isolation it invites expansion onto a broken slice, the outcome the gate exists to prevent. Now "re-run; fails → HALT as above, passes → continue, no checkpoint". The pinned expected string moved with it. Executor at 48,905 chars, 247 under the cap. Minor 4 — 2775 asserted two different current sizes for gsd-planner.md. The stale half is next's own text taken wholesale, so the contradiction was inherited; it now reads as a before-figure rather than a current one. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(#3299): cite the plan-md example by section, not by a drifting line Review round 7, Nit N-1. The 2775 ack fragment justified its one-line formatting with "matching docs/reference/plan-md.md:207's own example style". At head, :207 is prose; the one-line <verify><automated> example it means is at :222. The citation was accurate when written (77c2fda, f23205c) and drifted with a later merge of next. Re-pointed by section rather than by line — it has already drifted once, and the fragment's whole purpose is to be an accurate record — and the drift itself is recorded inline so the correction does not quietly overwrite what the earlier number said. Also narrows the changeset's "any task with gate=blocking-human" to "any tracer carrying gate=blocking-human" (found by Codex in the whole-PR pass). Golden rule 6 and the #3299 decision table both scope that gate to checkpoints and to the tracer feedback gate; the normal type="auto" branch never inspects `gate`, so the wider claim promised behavior the implementation does not have. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(#3299): answer fence-delimiter liveness by insertion, not replacement Review round 9. The round-8 fence-awareness fix was itself unsound, in the same class it was added to close. `operativeLineIndexes` detects operative lines by REPLACING each candidate with a throwaway predicate declaration and asking `parsePredicates` which survived. Sound for ordinary content lines. Not sound for a fence DELIMITER, which is exactly what the tracer-template selection passed it: deleting every ```xml OPENER leaves each matching closer to become an opener, and since `computeSkippedLineFlags` is a strict FORWARD state machine, fence parity inverts for the whole remainder of the document. Measured against the real file rather than argued: agents/gsd-planner.md has 3 live top-level ```xml openers — 0-based 180, 232, 262. operativeLineIndexes reported 180 and 262. Line 232, the "Task-level TDD" example, read NON-OPERATIVE — a wrong answer from a helper whose only job is that question. It passed only by parity coincidence, and one extra live example anywhere earlier flipped it to a false FAILURE blaming a decoy that does not exist: HEAD as-is | anchor 260 | openIdx 262 | ASSERTION PASSES +1 unrelated ```xml example | anchor 265 | openIdx 267 | ASSERTION *** FAILS *** Fixed by asking the question a way that perturbs nothing. `isOperativePosition` INSERTS a marker on its own line immediately before the candidate instead of replacing it. Insertion preserves every delimiter, and because the skip-state machine runs strictly forward, a line inserted at `idx` observes exactly the fence/comment state the candidate observes, with nothing but the marker between them — so marker-operative IS the candidate's position-liveness. The review's suggested direction (substitute a same-shaped opener that still opens a fence) cannot work here: the marker would then be inside the fence and would never parse as a predicate at all. Position-liveness is not content-liveness, so the helper also rejects a line that is entirely comment (`<!-- ```xml -->`), rather than leaving that to each caller's own shape test to happen to exclude. `operativeLineIndexes` now THROWS when its candidate regex matches a fence delimiter, so the unsound route cannot be reached again by a future caller rather than only being fixed at the one site that got it wrong. Verified with the same extra-example scenario above: with the fix, all 35 rows stay green. Teeth: reverting the call site to `operativeLineSet` turns the tracer-template row red on the new guard. The regression row pins both live openers (the second is the one the deletion route lost), the block-commented and same-line-commented openers, a line inside a fence, and re-checks both openers after unrelated lines shift above them. Only tests/tracer-bullet.test.cjs changes — no agent file is touched, so the 5-char gsd-planner.md and 19-byte gsd-executor.md headroom are unaffected. Verified: `npm run lint:ci` exit 0; full `npm test` 31307 tests / 31292 pass / 0 fail / 14 skipped, TMPDIR unset, against a freshly synced origin/next. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(#3299): guard the delimiter class, match the scanner, pin the assignment Codex full-PR review of #3390, run against the round-9 head. Three defects, two of them in the code that round added. 1. The mode-read pin survived the regression it exists to catch. `READ` matched the config-get substring only, so rewriting the shipped line as `IGNORED_MODE=$(gsd_run query config-get ...)` kept the row green while nothing defined HUMAN_VERIFY_MODE — the gate falls through to STOP and #3299 is back with the suite passing. The regex now requires the assignment. A lookahead after `end-of-phase` closes the other half: the bare prefix also accepted `--default end-of-phase-wrong`. Proven by mutation: renaming the variable in agents/gsd-executor.md now turns that row red, and did not before. 2. The round-9 fence-delimiter guard was a SAMPLE of the class, not the class. It probed a fixed list of five delimiter strings. `~~~xml`, ```json, `~~~~` and arbitrary info strings all walk past any list short enough to write down — the guard was added precisely because one such regex had already slipped through. Now matched against the lines the regex actually selects in the document, which cannot go stale and cannot miss a spelling nobody thought of. Four such spellings pinned as rows. 3. `isOperativePosition` disagreed with the scanner it delegates to. For `<!-- closed --> real content` it stripped the span, found surviving content, and answered "live". `computeSkippedLineFlags` skips an ENTIRE line whose trimmed text starts with `<!--`, balanced or not, before it considers fences at all. Verified directly against parsePredicates. It now applies the scanner's own rule instead of out-reasoning it. Latent for the present caller (its anchored ```xml shape cannot match a comment-prefixed line), real in general. Disclosed rather than fixed, and raised with the maintainer: the exact executor region pin ends before the second operative tracer-gate paragraph at agents/gsd-executor.md:327, which is only heading-checked — so contradictory later instructions could ship. How much of that file to pin is a call for its owner. Independently probed isOperativePosition across 19 edge cases before the review (line 0, CRLF, tab / 4-space / mixed " \t" indentation, 0-3 space fences, nested fences, ~~~ fences, info strings, bounds); all correct. That probe is what surfaced finding 2, which the review then confirmed from the other direction. Verified: `npm run lint:ci` exit 0; full `npm test` 31296 tests / 31281 pass / 0 fail / 14 skipped, TMPDIR unset. One caveat stated rather than smoothed over: in that run tests/planning-snapshot.test.cjs was truncated by concurrency after row A5 — 11 tests did not execute, which a 0-fail aggregate cannot show. Re-run in isolation it is 87 tests / 87 pass / 0 fail, and it is untouched by this change. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com> Co-authored-by: Tom Boucher <trekkie@nomorestars.com>	2026-08-23 18:43:53 -04:00
Behruz Nassre Esfahani	a44d513566	fix(#3712): confine in-process installs to a sandboxed HOME (#3725) ... * fix(#3712): confine in-process installs to a sandboxed HOME A runtime kind may declare a global `home` override resolved from os.homedir() rather than from the caller's configDir — codex's skills kind (`home: ".agents"`, ADR-1239 / #2088) is the only live case. Sandboxing configDir/targetDir does not contain it, and assertDestWithinConfigHome cannot see the class: that gate confines a destSubpath to whatever root it is handed, and here the root IS the escaped home. So an in-process caller that forgot to sandbox HOME wrote to, and pruned gsd-* entries from, the developer's REAL ~/.agents/skills. tests/agent-descriptor-parity.install.test.cjs's K1 loop did exactly that: it iterates every agents-kind runtime (codex included) with a sandboxed targetDir and an un-sandboxed HOME. Reproduced against a canary home on next @ adb46cdd8 — 71 gsd-* skill dirs deleted, a foreign `cloudflare` skill surviving, suite still exit 0. It is silent because the runtime's own config home is untouched, so the manifest keeps reporting a healthy install. FIVE writers resolve a kind `home` and then destroy under it. Three are reachable today — installRuntimeArtifacts, uninstallRuntimeArtifacts (install-engine.cts) and applySurface (surface.cts). Two are descriptor-dependent and guarded against a future descriptor change rather than a present escape: installOpencodeFamilySkills (behind the combined-family early return) and installAgentsKindStandalone. Those two are scoped to the single kind each destroys — passing the whole layout made codex's unrelated skills override trip a writer that never touches it. - src/test-home-guard.cts: refuse when a run under a test runner cannot be shown to have sandboxed HOME. NODE_TEST_CONTEXT (set by `node --test`) gates it, so installs outside a Node test context are untouched; GSD_TEST_MODE is unusable, as several candidate files including the offender never set it. Homes are compared by FILESYSTEM IDENTITY (st_dev + st_ino), not by pathname: path.resolve() resolves neither symlinks nor case, and realpath returns a canonical pathname that two routes to one directory can still disagree on (bind mounts). Verified on macOS/APFS — HOME=/users/<name> made the strings differ while naming the same directory, and the lexical form ALLOWED a write into the physical real home. FAILS CLOSED: a pair is "different" only when both identify, or one is definitively absent (ENOENT/ENOTDIR) while the other identifies; every other errno is "cannot tell" and refuses. Only when neither home identifies is a marker consulted, and it carries the sandbox PATH and must equal the home in effect — a boolean checked first let an ambient or stale value disarm the guard. - helpers: promote sandboxHome() out of its two byte-identical private copies, which is also what makes them record the sandbox; the three withFakeHome() helpers record it too. The marker NAME is duplicated as a bare string rather than required from the compiled guard, keeping helpers.cjs's documented no-built-lib-at-import-time contract; a test pins the two together. - agent-descriptor-parity: sandbox HOME across the K1 loop. - helpers-process-isolation: #3156's canary asserts on <home>/.gsd only, and its `--cursor --local` spawn cannot reach `.agents` at all, so an assertion added there would pass with all confinement removed. Add a discriminating row — a `--codex --global` spawn against a seeded ambient home — which also asserts the runtime still declares the override. Its check is a sampled inventory (dir names + each SKILL.md), not a tree compare. - install-write-confinement: predicate rows through the deps seam, covering the symlinked HOME, ambient and stale markers, and each sameDirectory branch (both-identify, one-absent, neither-identifiable), plus wiring rows that drive the REAL entrypoints so deleting a guard call site is red. Verified: guard fires end-to-end against a real un-sandboxed HOME (exit 1, zero deletions); the case-variant fail-open reproduced on APFS before the fix and refuses after; K1 file 29/29 green with skills intact; mutation-tested — each of the three reachable call sites, lexical-only comparison, and treating an unknown errno as "absent" each take exactly one row red, with every mutation echoed back; the process-isolation row negative-controlled by reverting installerEnv to its pre-#3156 leak (16/0 -> 13/3); a full npm test leaves ~/.agents/skills at 71. Stated residual: the two descriptor-dependent writers have no wiring test, because no runtime declares a `home` override on those kinds and neither can be exercised without inventing a descriptor. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#3712): add changeset Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3712): let a sandbox nested inside the real home through the guard All six Windows shards of #3725 failed on legitimately sandboxed destinations. On Windows os.tmpdir() is %LOCALAPPDATA%\Temp — inside the user's home — so every sandbox a test creates is a descendant of the real home, and "does this land inside the real home?" answers yes for the safe case and the dangerous one alike. POSIX conceals this: /tmp and /var/folders both sit outside $HOME. Add the missing conjunct: a destination inside the real home is allowed only when it also sits beneath a HOME that was sandboxed away from the passwd home. Both halves are required — dropping the first re-admits a plain un-sandboxed install, and dropping the second decays into the "is HOME sandboxed?" check the module rejects, which a layout resolved before the sandbox walks straight through. Each is mutation-proven by a row that goes red without it. Also covers the two fail-closed branches of the new exemption, which survived mutation to `true` with the suite green, and avoids `<user>` in a docblock — the prompt-injection scanner reads it as a delimiter tag. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(#3712): close three writer/rollback gaps found reviewing the whole PR Cross-AI review of the full PR (not just the round's delta) surfaced three ways the guard could still be defeated: - The nested-sandbox exemption trusted the SPELLING of a destination. With HOME sandboxed to a directory inside the real home — legitimate on Windows — an aliased `.agents` (symlink, junction, subordinate bind mount) beneath it redirected an allowed path into the real home. Decide containment on the path the write RESOLVES to: walk up to the nearest existing ancestor, canonicalize, re-append the tail. - `migrateLegacyDevPreferencesToSkill` is a SIXTH writer that resolves a skills-kind `home` override. It creates rather than prunes, which is why it was missed, and `_runLegacyInstallMigrations` runs it before `installRuntimeArtifacts`' own assertion. Guarded, scoped to that kind. - Worst of the three: `bin/install.js` snapshots the resolved skills root before installing, and its outer catch rolls back by deleting and recreating every snapshotted `gsd-*` directory there. The guard's own throw landed in that catch, so refusing an un-sandboxed codex install provoked exactly the mutation the guard exists to prevent. Refusals are now marked and rethrown without rollback — nothing was written, so there is no partial install to undo. Every other error still rolls back. Also carries the sandbox marker into `installSpawnEnv`, so spawned installers are not refused on passwd-less CI images, and corrects three claims that no longer hold: "every writer" (six, and named), the unconditional "fails CLOSED" (the passwd-less marker branch is a deliberate weakening, and TOCTOU is out of scope), and the assertion that Windows os.tmpdir() is always %LOCALAPPDATA%\Temp (Node honors TEMP/TMP). Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * docs(#3712): name the guard's two limits instead of overclaiming Round-2 review found the prose had drifted ahead of the code. Corrected, with no behavior change: - The module still said FIVE writers; there are six, and the sixth is now named along with why it was missed (it creates rather than prunes) and why it carries its own assertion (it runs before the main one). - The canonicalization docblock listed subordinate bind mounts among the aliases it closes. It does not close them: a bind mount is not a link, so realpath keeps the mount-point spelling. `sameDirectory` already recorded that limit; the new helper now inherits it explicitly rather than contradicting it. Closing it needs mount-table introspection. - "FAILS CLOSED" was unqualified while the passwd-less marker branch is a deliberate weakening — with no passwd entry, nothing can contradict a marker naming the real home. - "Refuses BEFORE any write" was too broad: legacy install migrations run ahead of the layout-driven ones, which is exactly why the two rollbackInstallerMigrations() calls still execute before the rethrow. Only the codex skills-root rollback is skipped, and that is the only _codexPreConfigRollback() call site — applySurface is never called from bin/install.js and uninstall cannot reach it. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(#3712): sandbox HOME in the opencode-family home-override parity rows The last two Windows failures, and the same platform asymmetry in a different disguise. This row drives a skills-kind `home` override on purpose — precisely what the guard polices — but relied on the override temp dir happening to sit outside the real home. It does on POSIX (/tmp, /var/folders); on Windows os.tmpdir() is under %USERPROFILE%, so the guard correctly refused and only Windows went red. Declare the sandbox instead of depending on the platform: HOME becomes the override itself, which is the home the call writes under. This is the fix the guard's own message prescribes, applied to the test rather than to the guard. Both failing Windows shards fail on exactly these two rows and nothing else; every other shard is green. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(#3712): make sameDirectory answer NO when it cannot tell Review Major 1. sameDirectory()'s only caller is the passwd-less marker branch, which reads a `true` as permission to PROCEED: if (marker && sameDirectory(marker, osMod.homedir())) return; The fallthrough returned `true` whenever neither side identified — two absent paths, or two stats failing EACCES/EPERM/EIO on a locked-down host — on the reasoning that "cannot tell" should make the caller refuse. That reasoning was inverted with respect to this caller: it turned the passwd-less escape hatch into an unconditional bypass for any marker value at all, on precisely the hosts the fallback exists to serve. Only two things now answer yes: one resolved pathname, or two readable identities that match. Restoring the old fallthrough takes the new row red. Also from review: - Major 2 asked whether st_dev/st_ino discriminate directories on Windows, where Node derives them from BY_HANDLE_FILE_INFORMATION. The whole guard rests on that primitive, so assert it rather than argue it: a row comparing two distinct temp directories, and one directory reached by two spellings. It runs on every platform in the matrix, so Windows answers the question itself. - Minor 1: the refusal now names the real home it compared against, not just the destination it refused. That is the one fact needed to tell a true positive from a false one, and its absence is what made the Windows case a CI-log dig rather than a glance. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(#3712): refuse a HOME that merely spells the real home more widely Review round: one Blocker, four Minors, a Nit. N2 (the one with teeth) — a destination's ancestor chain is linear, so "inside the real home AND inside the effective HOME" admits two arrangements, not one. The intended `effectiveHome ⊂ realHome` is the Windows temp shape; `realHome ⊂ effectiveHome` — HOME at /Users, /home, C:\Users — is not a sandbox at all, it is the real home reached by a wider spelling, and it was exempting a stale destination pointing straight at ~/.agents. Third conjunct added; the docblock no longer claims two conditions suffice. Removing the conjunct reds the new row and nothing else. N4 — the migration guard resolved its OWN layout, and without capabilityRegistry, so a registry-dependent descriptor could make it vouch for a path the migration does not write: a guard reporting safe while the unsafe write proceeds. It now guards the destination already resolved by _resolveDevPreferencesSkillTarget, keyed on `installRoot !== targetDir` — which is exactly the condition under which a `home` override was declared, read off that same result. N1 — CONTEXT.md gains the Test Home Guard Module glossary entry that contributor-standards.md requires of a new Module. Not CI-enforced, so green CI was never evidence it was met. N3 — the docs/INVENTORY.md row was misfiled between install-fs-adapter and install-model-override-resolver; the table is alphabetical and the manifest already had it right. Moved, and its text now names six writers and the third conjunct. N5 — applySurface's signature docblock was two parameters stale; this PR added the second of them. N6 — the duplicated rollbackInstallerMigrations() adjacent to the new rethrow: two identical consecutive calls, not two phases. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(#3712): guard the sixth writer, and close two false-ALLOW paths Review round 3 (NEW-1, NEW-2) plus three defects Codex found in the whole-PR pass, each reproduced before it was fixed. NEW-1 — migrateLegacyDevPreferencesToSkill called the guard with no `deps`, so it bound real os/process.env and could not be wiring-tested the way the other three reachable writers were. It now takes the same optional `deps: { os?, env? }` tail parameter. The wiring block gains the missing fourth row, and a fifth pinning the ALLOW half; the test file's header docblock said "FIVE writers ... the three reachable today", contradicting the six/four statement this PR already put in src/test-home-guard.cts, CONTEXT.md, docs/INVENTORY.md and the changeset. Both directions of the guard's condition now fail a row when broken — previously neither did. NEW-2 — derivesFromSandboxedHome's docblock claimed "THREE conditions are required, and no two of them suffice". False for {2,3}: isInside is reflexive, so whenever conjunct 1 fires conjunct 2 already returns false on its own. Reworded as a fast path, which is what it is. Codex 1 (false ALLOW) — on a host with no readable passwd entry the marker branch returned as soon as the marker matched the effective HOME. That attests a caller sandboxed HOME and says nothing about where an already-resolved destination points, so a layout captured before sandboxHome() — still naming the real ~/.agents — was waved straight through: the same stale-layout shape the primary branch refuses by design. The marker must now identify AND contain every destination. Codex 2 (false ALLOW) — `installRoot !== targetDir` was the stand-in for "the skills kind declared a home override". The two are not equivalent: the inequality is false when the override resolves onto targetDir itself, which is exactly a configDir of $HOME/.agents. The guard was skipped and SKILL.md written into the real home under a test runner. _resolveDevPreferencesSkillTarget now reports hasHomeOverride off the same resolution instead of inferring it from two paths. Codex 3 (prose) — the shared refusal message claimed every guarded writer prunes; the migrate writer only creates. The changeset headline claimed in-process installer calls can no longer reach the real home, which is wider than the guard: writeNonClaudeDefaults still writes ~/.gsd/defaults.json through os.homedir(). INVENTORY's and CONTEXT's fail-closed sentences omitted sameDirectory's pathname-equality shortcut. All four narrowed to what the code does. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(#3712): let the sandbox marker follow an overridden HOME Found by Codex in the whole-PR pass. installSpawnEnv spreads `overrides` last so an explicit HOME wins — deliberate, and its docblock tells callers needing per-spawn isolation to pass their own { HOME, USERPROFILE }. But the #3712 marker was set before that spread, so such a caller got HOME=<theirs> and marker=<helper default>. On a host with no readable passwd entry the guard compares the two and refuses a legitimately sandboxed spawn — tests/install.test.cjs:7143 and install-shared.cjs's own runInstaller both take that path. The marker is now derived from the final HOME unless the caller supplied one explicitly. The contract test asserted HOME after an override but not the marker, which is why it stayed green. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * docs(#3712): name the shipped guard condition, not the deleted one Review round 4 of #3725. Two artifacts this PR adds still described `target.installRoot !== targetDir` in the PRESENT tense as the live guard condition on `migrateLegacyDevPreferencesToSkill`. The shipped condition is `runtime && target.hasHomeOverride` (src/install-engine.cts:510). This is not ordinary doc drift. The named condition is the exact false-ALLOW the previous round closed: a `home` override resolving onto `targetDir` — a configDir of `$HOME/.agents`, which is where codex's override points — makes the inequality FALSE while the override is declared, so the guard was skipped. A maintainer reading CONTEXT.md:290 as authoritative would believe the guard still skips that case. - tests/install-write-confinement.test.cjs — the ALLOW-half row's comment. Its "teeth" rationale is unchanged and still correct as written. - CONTEXT.md:290 — the Test Home Guard Module glossary entry, a documented PR gate. Now states the condition and names the inequality only as what it is NOT, with the reason. The three surviving mentions of the inequality are all past-tense or negated (src/install-engine.cts:448, :508 and the sibling test comment at :3698) and are correct as they stand. Verified: `npm run lint:ci` exit 0; full `npm test` 31327 tests / 31312 pass / 0 fail / 14 skipped, run with TMPDIR unset. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(#3712): canonicalization fails closed, matching identify's errno split Codex full-PR review of #3725, run against the round-4 head. `resolveThroughLinks` caught EVERY realpathSync error and fell back to `path.resolve(dest)` — the lexical spelling. That inverts the function's own purpose. An aliased `<sandbox>/.agents` that cannot be canonicalized keeps its sandbox spelling, satisfies the nested-sandbox exemption at :227, and the write is ALLOWED into the real home — the exact escape this walk exists to close. The module documents that it fails CLOSED with ONE named exception (the marker branch); this was a second, unnamed one. Split by errno, and deliberately by the SAME split `identify` already draws rather than a second policy in one module — both answer "does this path exist as named?", so they must not disagree: ENOENT / ENOTDIR -> walk up. The ordinary case: a fresh install resolves a destination nothing has created yet, so realpath fails on the leaf and on every not-yet-created ancestor. Refusing here rejects every install. anything else (EACCES, EPERM, ELOOP, EIO) -> refuse. The component exists but cannot be resolved, so the guard cannot tell where the write lands. Three rows in the predicate block, beside the other aliasing rows: - a symlink CYCLE in the destination path (ELOOP) -> REFUSE - a destination that does not exist yet (ENOENT) -> ALLOW - a component behind a regular file (ENOTDIR) -> ALLOW Teeth checked against the artifact the test loads, not the source: reverting the condition to the swallow-everything shape in the compiled test-home-guard.cjs turns row 1 — and only row 1 — red. The ENOTDIR row caught a stale build during development, which is the point of asserting on the compiled file. CONTEXT.md and the resolveThroughLinks docblock both record the new behaviour, so this does not repeat the prose-vs-code drift the round-4 finding was about. The changeset's existing scope sentence now bounds "six writers" to the `installRuntimeArtifacts` call tree and names `cmdGenerateDevPreferences` — which resolves the same codex `home` override through `getGlobalSkillsBase` and writes SKILL.md beneath it unguarded. It has no in-process caller today (its only direct require-and-call is a spawnSync with HOME sandboxed), so it is latent rather than live, and whether it belongs in this PR is raised with the maintainer rather than decided here. Verified: `npm run lint:ci` exit 0; full `npm test` 31330 tests / 31315 pass / 0 fail / 14 skipped, TMPDIR unset. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-23 18:27:02 -04:00
Tom Boucher	004e9dd741	fix(#3007): resolve Codex reasoning effort per model and make every clamp visible (#3765) ... * test(#3007): failing-first suite for per-model Codex effort capability RED by construction. Binds to behavior renderEffortForRuntime does not yet have: an optional third `model` argument, a per-model advertised-level table, `max` passing through instead of clamping to `xhigh`, `minimal` clamping to `low`, `ultra` rejected outright, and clamp visibility (`requested`/`clamped`/ `reason`) so a downgrade is legible from resolver output rather than silent. Two of these pin defects that exist on next today: - `max` is discarded. Both Codex models whose catalog entries are retrievable (sol, luna) advertise `max`; GSD clamps it to `xhigh` and reports nothing. - `minimal` is emitted to a model that refuses it. providerPresets.openai. haiku.low pairs gpt-5.6-luna with reasoning_effort "minimal", and luna's advertised floor is `low`. GSD is sending a value into a document Codex itself validates. The parity test is what pins that fixed, and it names the offending path/model/effort when it trips. Also corrects tests/model-resolver.test.cjs:351, which asserted renderEffortForRuntime('codex','max').value === 'xhigh' -- the defect pinned as though it were a contract. ADR-443 recorded "Codex has no max" as fact and it was true when written; Codex has since added both `max` and `ultra`. That is a stale premise, so the assertion is corrected here rather than worked around. The property test asserts the invariant the whole change exists for: a rendered effort is always a level the target model actually advertises, or an explicit rejection. There is no third outcome. * fix(#3007): resolve Codex effort per model, and make every clamp visible Codex declares supported_reasoning_levels per MODEL and validates against it, so a single per-runtime capability set cannot be right for all of them. GSD's was wrong in both directions at once. `max` reaches Codex now. ADR-443 recorded "Codex has no max" as fact and clamped max -> xhigh on that basis; it was accurate when written, and Codex has since added both `max` and `ultra`. Every Codex model whose catalog entry is retrievable advertises `max`, so the clamp was discarding a level the provider supports, silently, on the most-used path. `minimal` stops reaching Codex. No Codex model advertises it -- both retrievable entries floor at `low` -- yet providerPresets.openai.haiku.low paired gpt-5.6-luna with reasoning_effort "minimal". GSD was writing a value the receiver validates and refuses into a file the receiver reads. Being unconservative in what you send is the half of Postel's rule with no defensible reading, so that preset is corrected and a parity test pins it. `ultra` is refused rather than laddered. Codex's own catalog calls it "Maximum reasoning with automatic task delegation": at ultra, effective_multi_agent_mode returns Proactive and Codex spawns sub-agents on its own initiative, underneath GSD's orchestration rather than inside it (#2167). It is a mode switch, not a reasoning depth, so it is not added to the universal ladder -- which stays provider-agnostic by ADR-443's design -- and it is rejected even for gpt-5.6-sol, which does advertise it. Clamping it down to `max` was considered and rejected: that silently discards what the user actually asked for. Clamping is now visible. RenderedEffort carries requested/clamped/reason and resolve-execution surfaces them. The previous table clamped correctly but invisibly, so a user asking for `max` on Codex had no way to find out they were getting `xhigh` -- exactly the failure mode the robustness principle's modern critique warns about, and why "be liberal" has to mean "liberal and loud". Also closes a latent trap found while reviewing the implementation: the clamp-up loop walks the ladder upward, and for a future model advertising `ultra` but not `max` it would have selected `ultra` as the clamp target -- re-entering by the back door the mode the rejection above exists to keep out. A clamp may never produce a value that a direct request for that value would refuse. Unreachable with today's catalog, which is why no test caught it; a test now asserts the invariant directly. Signature stability is preserved: the third `model` argument is optional and the two-argument form still resolves, against the family baseline. That form's BEHAVIOR does change for `max` and `minimal`, and it must -- keeping the old answer would have fixed the defect only where a model happened to be threaded through and left it live everywhere else. tests/model-resolver.test.cjs:351 asserted the defect as if it were a contract and is corrected here rather than worked around. * fix(#3007): close every review finding on the Codex effort alignment Two isolated reviewers, correctness and security. Both found the same two blockers, and the per-model work was inert on every surface that matters until this commit. BLOCKER — resolve-execution never passed the model and discarded the clamp. cmdResolveExecution called the two-argument form and emitted only effort_rendered/effort_param/effort_propagation, so the per-model table was unreachable from production code (tests were its only caller) and requested/ clamped/reason were computed and thrown away. Requested outcome 3 names "the effective rendered effort in resolver output" specifically, so the feature was unmet on the exact surface the issue asks for. Now passes the resolved model and emits effort_requested / effort_clamped / effort_clamp_reason, flat, matching the existing key convention rather than introducing a nested object. BLOCKER — the docs described output that did not exist. CONFIGURATION.md showed a nested {"effort": ...} sample; the real result is flat and those keys were absent entirely. A reference doc asserting a JSON path a reader can copy is worse than no doc. Corrected against the actual emitted key set. MAJOR — the argv channel still shipped both original defects. EFFORT_ARGV.codex kept minimal in its supported set and still clamped max down to xhigh, so the invocation-time and install-time channels disagreed about the same runtime's capability: --host codex with max emitted xhigh while the generated TOML said max. This is the repo's documented generative-fix-divergence class, so both tables now cross-reference each other and a parity test fails if they ever diverge again. MAJOR — malformed catalog data failed OPEN and could crash the CLI. A null _baseline became an EMPTY Set that is nonetheless truthy, so the nullish fallback never fired and every effort rendered as null. And a non-array value made the Set constructor throw at module load — model-catalog.cjs is required across the whole CLI, so one bad JSON value killed every command, not just codex effort. Guarded on size and filtered to array values; both degrade to the hardcoded baseline. MAJOR — value widened to a nullable string with two consumers left behind. runtime-artifact-conversion passed it straight into injectEffortFrontmatter (a null effort key in generated frontmatter); install-effort-resolver still declared a non-nullable return, a structural lie that silently defeated null checking. Both corrected, both omitting the key on null — the same posture as 'inherit', where omission means "follow the host default". MAJOR — the per-model table is inert today, and the docs now say so. All three shipped models advertise the same usable range and ultra (sol's only differentiator) is rejected for every model, so no observable output differs by model. The table stays because Codex declares capability per model and the sets are free to diverge — a single per-runtime assumption is precisely what went stale and produced this issue — but overselling it as a visible per-model feature would have been the same class of error as the doc blocker above. Tests: three passed under a full revert and are strengthened rather than deleted, since each guards a real contract (#3533's inherit rule, the undeclared-host rule, off-ladder handling) — they now also assert the clamp-visibility fields, which only exist after this change. The fast-check property is kept for its shrinking, and a deterministic nested loop over the full cross-product now sits beside it so coverage is exhaustive rather than sampled. Also folded in earlier: bin/install.js generated the Codex TOML with the two-arg form and would have written a literal null reasoning effort on the ultra path; CONTEXT.md's Model Catalog Module glossary entry now records CODEX_MODEL_EFFORT. The installer defect was found by the co-change gate, not by a reviewer — install.js is a historical co-change partner of model-catalog.cts that this diff had not touched. * test(#3007): correct assertions that pinned Codex's stale effort premise Thirteen pre-existing tests encoded "Codex has no max" as fact and failed on the shipped commit. Every one is a stale pin, not a defect: each was probed against the built module before its expectation was changed, and none failed for a reason other than this premise correction. Kept as its own commit per CONTRIBUTING — a test-fixture correction made stale by a production change must not ride inside another commit, because the release-sdk hotfix cherry-pick filter routes by subject prefix and a correction buried under the wrong prefix ships a half-state (v1.42.3, #3621). The most valuable one was tests/model-resolver.test.cjs's cross-provider validity invariant, which hardcoded the Codex enum as `minimal|low|medium|high|xhigh` and failed with "real API would 400". That message is now false in both directions: Codex accepts `max`, and rejects `minimal`, which no model advertises. The enum is corrected to `low|medium|high|xhigh|max` and the guard is kept intact — it is exactly the "would the real API refuse this" check worth having, and it was right to fail here. It simply carried the stale fact in its own fixture. Test NAMES were corrected alongside their assertions wherever the name asserted the old behavior — "max is Anthropic-only", "max clamps to xhigh", "minimal passthrough". A renamed test that still claims the old thing is worse than a failing one, and a green test whose name states a falsehood is how the next reader inherits the wrong premise. Both channels are covered: install-time (renderEffortForRuntime, and the generated .toml in install-runtime-artifacts) and invocation-time argv (effort-surface-axis). They were deliberately brought into agreement in this change, so their assertions had to move together. Each site carries a #3007 comment recording that Codex gained max/ultra and that capability is declared per model, so a future reader can tell this was a deliberate premise correction rather than a test bent to fit an implementation. * test(#3007): separate the effort-precedence case from the clamp case The previous stale-assertion pass over-corrected one test. It saw `effort: { default: 'max' }` on codex expecting `effort_rendered: 'xhigh'`, assumed the xhigh came from the max→xhigh clamp #3007 removes, renamed it to "max passes through" and changed the expectation to `max`. The remote runner disagreed. Reproduced against the real CLI: with that config and `gsd-planner`, the resolver emits `effort: "xhigh"`, `effort_requested: "xhigh"`, `effort_clamped: false`. The xhigh is produced by effort-resolution PRECEDENCE — gsd-planner is heavy/opus tier and its routing-tier default outranks `effort.default` — so `max` never reaches the renderer at all. The test says nothing about clamping and never did; it only looked like a clamp pin because both mechanisms happened to yield the same string. Restored to `xhigh` and renamed to say what it actually tests. It now also asserts `effort_clamped === false` and `effort_requested === 'xhigh'`, which is what makes it impossible to mistake for a clamp pin again: those two fields prove the value is what the resolver produced rather than something the renderer downgraded. Before #3007 there was no way to tell the two apart from the output — which is precisely why the previous pass could not tell them apart either. Added the test that was actually missing: `effort.agent_overrides`, which outranks the tier default, so the requested level genuinely reaches the renderer and `max` survives to `effort_rendered` end-to-end through the real CLI. Verified by probe before asserting. One test now pins the precedence rule and the other pins the #3007 behavior, and neither can be read as the other. That the clamp-visibility fields are what resolved this is a small argument for having added them. * chore(#3007): backfill changeset pr number to 3765 * test(#3007): put model-catalog under the mutation gate The Stryker shard showed as `skipping` on this PR despite the diff rewriting model-catalog's effort logic. That was legitimate, not a detection bug: `model-catalog` was never in scripts/mutation-matrix.cjs's COVERED map, so the whole module — including everything #3007 touches — sat entirely outside mutation scoring with has_work "false". Registered, with a dedicated spawn-free surface. tests/model-catalog.unit.test.cjs is new: 44 in-process tests, no runGsdTools, no child process, no filesystem, no temp dirs. That shape is not stylistic — it is the #2790 precedent this file already documents. Stryker's command runner treats a whole `node --test <file>` invocation as ONE test costing whatever its slowest case costs, and re-runs it per mutant, so pointing a shard at tests/model-resolver.test.cjs (which uses runGsdTools throughout) would reproduce exactly the 15-minute shard-cap cancellation #2790 hit. The integration file is unaffected and keeps running in full in the normal test job. Coverage spans the module rather than only the diff, because the score is measured over the whole file: effort rendering across every model and ladder level in both channels, the prototype-chain host guard, the exported enums and maps, isAnthropicFlavoredModel's provider namespacings, the profile projections, nextTier, and mergeEffortTierDefaults. The last two were nearly left out and are worth naming — every uncovered exported function is score given away, and mergeEffortTierDefaults turned out to have a genuinely interesting contract (#3531: a partial override merges over the built-ins rather than replacing them, and isValid gates the VALUE, not the tier name, so an unknown tier key is still merged in). Every expectation was probed against the built module before being asserted. minScore is 1 and that is a PLACEHOLDER, flagged as such in the registry comment. Floors in this repo are measured, not chosen — the existing entries sit at 94, 75 and 56 — and they can only be measured in CI, because mutation shards run `node --test`, which is hard-blocked locally. The first CI run on this branch reports the real number and the floor gets ratcheted to it before merge. A placeholder of 1 reaching `next` would make the gate decorative: it would pass whether or not a single mutant is ever killed. Note the target is "never regress from measured", not a fixed 80 — planning-inspect sits at 56 and is documented as an accepted ratchet candidate. * test(#3007): bootstrap model-catalog's mutation floor legally The placeholder floor was structurally illegal and the remote run said so. tests/mutation-matrix-ratchet.test.cjs guards the guard: every COVERED module must carry a matching RATCHET_BASELINE entry in the same diff, minScore must EQUAL that baseline, and it must be at least 50. `minScore: 1` failed all three. That is the ratchet working exactly as intended — a floor nobody can satisfy accidentally is the point of it. Bootstrapped at 50 in both places. Fifty is not a measured score and the comment says so plainly: it is the minimum the guard permits, and it coincides with Stryker's own configured `break` threshold, so it is the lowest legal starting point for a module that has never been measured. It still must be ratcheted to floor(measured) - 1 before this PR merges. Also corrected a real defect in the file's own instructions. "HOW TO UPDATE" step 1 read "Run the per-module Stryker shard locally" — which cannot be done here, and which the same file contradicts eighty lines further down, where the #2790 scores are recorded as "not a local run; mutation shards run `node --test`, hard-blocked in this repo's local environment". stryker.config.mjs confirms the command runner invokes `node --test` once per mutant, and .claude/hooks/block-local-node-test.sh denies exactly that. So the documented first step sends the next contributor at a wall. Rewritten to describe the path that works — push, read the measured score off the CI shard, then set the floor and its baseline together in one diff — and to say why local measurement is not available, so nobody rediscovers it the slow way. GOODHART SAFETY is untouched. The two-step is inherent to the environment rather than a shortcut: a floor cannot be measured before the first CI run exists, and the guard rightly refuses to accept an unmeasured one below its minimum. * test(#3007): ratchet model-catalog's mutation floor to its measured score The shard ran in CI and reported 59.62% — 248 mutants killed, 168 survived, no timeouts, no errors (run 32605073352, job 97108869486). Floor set to 58 per this file's own rule, minScore = floor(measured) - 1, which is the same arithmetic every sibling entry used: 57.03 to 56, 76.58 to 75, 95.65 to 94. Both halves moved together, because the ratchet guard asserts minScore equals its RATCHET_BASELINE entry and would reject them drifting apart. The spawn-free unit surface is vindicated by the clock: 57 seconds, against a 15-minute shard cap and a 9m46s frontmatter shard in the same run. That was the whole reason for creating tests/model-catalog.unit.test.cjs rather than pointing the shard at tests/model-resolver.test.cjs — #2790 recorded shards being CANCELLED at that cap when they targeted a runGsdTools-heavy integration file. The registry comment is rewritten rather than deleted. It previously warned that the floor was provisional and must not ship that way; leaving that text next to a measured floor would make the file lie in the other direction. It now records the measurement the way the sibling entries do, including that 59.62 sits below TARGET (80) and is therefore a ratchet candidate like planning-inspect at 56 — comfortably clear of its own floor with real room to grow. Raise it as the tests improve; never lower it. Worth stating plainly: 168 surviving mutants is not a clean bill of health. It is an honest floor for a module that had NO mutation coverage at all an hour ago, and it is now pinned so it cannot silently regress. --------- Co-authored-by: sim <sim@local>	2026-08-22 20:51:55 -04:00
Tom Boucher	b6977d9d11	fix(#3762): enforce docs/INVENTORY.md roster rows, backfill 32 gaps (#3766) ... * test(#3762): failing-first roster gate for docs/INVENTORY.md rows Anchors the human half of the inventory-drift rule: every entry in docs/INVENTORY-MANIFEST.json must have a hand-written row in docs/INVENTORY.md. Expected RED on this commit -- next carries 32 unrostered surfaces, which is the defect the gate exists to catch. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3762): enforce docs/INVENTORY.md roster rows, backfill 32 gaps docs/INVENTORY.md calls itself the authoritative roster of every shipped GSD surface, and CLAUDE.md's inventory-drift rule requires both a roster row and a manifest regen. Only the manifest half was anchored, so a PR could ship a surface, regenerate the manifest, omit the row, and stay green -- as PR #3758 did with gsd-core/references/planner-coupling.md. Adds the roster half to tests/inventory-manifest-sync.test.cjs, backed by a pure matcher in tests/helpers/inventory-roster.cjs, and backfills the 32 surfaces already missing rows on next. Fixes #3762 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test(#3762): harden roster matcher against fenced blocks and trim exports Skips fenced code regions when splitting level-2 sections so a documented '## ' example inside a fence cannot truncate a family section (false red) or contribute a phantom row (false pass); makes the heading pattern linear rather than a backtracking lazy match; narrows the module surface to the three names the gate consumes. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3762): close two false-RED gaps found in orthogonal review Indented headings: CommonMark permits an ATX heading to carry 1-3 leading spaces, and a ^##-anchored pattern read such a document as having no family sections at all -- reporting all six missing, a structural red for zero real drift. Reproduced, then fixed and pinned. Link-wrapped cells: a row written as [`x.md`](../x.md) was not recognized, though docs/INVENTORY.md already uses that form elsewhere. Unwrapping now peels a whole-cell link and a whole-cell code span, and only layers that wrap the cell entirely -- a file mentioned mid-prose is still not a row, so the false red is not traded for a false pass. Adds a second fast-check property over the commands source-link rule, the one family whose matching rule differs from the other five. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test(#3762): boundary trio for the fence-marker length limit RULESET.TESTS.boundary-coverage — the matcher's only numeric limit is the fence marker's {3,}. Exercises 2 (inline markup, swallows nothing), 3, and 4 characters, for both backtick and tilde delimiters. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore: remove stray pwned_cmdsub injection-test canary from the repo root A zero-byte file committed by 0e6fa2e2c (#3124) while remediating the #3118 command-substitution injection — the marker a test wrote into cwd to prove a substitution had NOT executed, left behind when the run ended. Nothing in the tree references it (verified by Grep across the repo), package.json's files array excludes root-level files so it never shipped, and lint-removed-but-needed confirms no surviving reference. Found while auditing the repo root for #3762; fixed inline rather than deferred, per CLAUDE.md's no-deferrals rule. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs: backfill changeset pr number to 3766 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-22 18:24:25 -04:00
Tom Boucher	2f86278b5e	fix(#3003): opt-in mechanism for intentional deletions in worktree.cleanup-wave (#3757) ... * test(#3003): failing-first suite for declared deletions in cleanup-wave Binds the guard's opt-in before it exists, so the suite is RED against next. The rows that carry the weight are the over-authorization set: a directory declaration must not authorize its children, a glob declaration must authorize nothing, and a declaration must not act as a string prefix of another path. Each of those BLOCKS, and each would PASS under a prefix, glob, or startsWith matcher — which is how a path list quietly degrades into the boolean opt-in #3003 explicitly rejected. The glob row matters most: declaredScopePrefix already returns null ("matches everything") for a glob-leading pattern, correct for the advisory it serves and catastrophic for a gate. Also pinned: a failed deletion check blocks on its own reason rather than being filtered into a pass; the block detail names only the undeclared residue so the operator is not misdirected by paths that were fine; an entry with no declaration blocks exactly as before; junk and non-array declarations do not authorize; and a blocked entry still isolates rather than aborting the wave (#2852, which must stay fixed). Two advisory rows cover an interaction found while designing: git diff --name-only includes deleted paths, so without unioning the declaration into the #2596 scope check, authorizing a deletion would raise SCOPE_OUT_OF_DECLARED against the very path just authorized. A seeded property states the whole invariant the three over-authorization rows sample: a deletion merges iff its normalized path is in the declared set. * feat(#3003): declared deletions opt-in for the cleanup-wave guard The deletions guard blocked the merge-back of any executor branch whose diff removed a file, with no way to say a removal was intended. A plan that folded one test file into a sibling could not be merged by the tool meant to merge it, forcing a manual --no-ff outside the tool -- strictly less safe than what the guard protects against. A plan now declares removals in its own frontmatter (files_deleted), and that list rides the same path files_modified already travels: plan-document parse -> phase plan JSON -> the per-plan worktree gate -> record-agent/create --deletions -> declared_deletions on the manifest entry -> the guard. The guard blocks only the deletions NOT in that list. A path list rather than a boolean, per the pinned decision: a boolean disarms the guard for the whole entry, so an unexpected deletion riding along with a declared one would pass unnoticed. Matching is exact after the module's shared normalizer -- never a prefix, never a glob. Both would let one declaration authorize a whole set, which is the mass-deletion accident the guard exists to catch. That also means declaredScopePrefix is deliberately NOT reused here: it returns null ("matches everything") for a glob-leading pattern, which is right for the advisory it serves and would silently disarm a gate. The block detail now carries only the undeclared residue, so an operator is not sent looking at paths that were fine. A failed deletion check still blocks on its own reason and is never filtered into a pass. A blocked entry still isolates rather than aborting the wave (#2852). The #2596 scope advisory unions the declaration into its declared set -- git diff --name-only includes deleted paths, so without that, authorizing a deletion would immediately warn that the same path was out of declared scope. Optional and additive throughout: files_deleted is absent from PLAN_REQUIRED_FIELDS, a manifest entry without declared_deletions keeps the original unconditional block, and omitting --deletions leaves the on-disk entry shape untouched. Supersedes the spent #2856 emitted-drift ack entry for execute-phase.md, the same supersede that entry performed on #3370 and #3370 on #3324. * fix(#3003): wire --deletions on every dispatch surface, not just one Review found the feature inert on two of three dispatch paths. execute-phase.md (harness inline) passed --deletions, but the orchestrator-worktree path (executor-isolation-dispatch.md, worktree.create) and the Fleet-parallel batch path (capabilities/claude-orchestration/fragments/execute-wave-pre.md, worktree.record-agent) still passed only --files. A plan declaring files_deleted would have merged on one path and been blocked on the other two -- the exact bug #3003 exists to fix, left unfixed where most of the isolation actually runs. Worse, per-plan-worktree-gate.md already claimed --deletions was passed 'on the same worktree.record-agent / worktree.create calls', which was false for both untouched sites. A doc asserting coverage that does not exist is how a gap survives review. All four surfaces now pass the flag, verified by sweeping every .md under gsd-core/, capabilities/, commands/, skills/ and agents/ that invokes worktree.record-agent or worktree.create: each one that passes --files now also passes --deletions. The isolation-dispatch note explains why this flag, unlike --files, is not advisory -- omitting it does not skip a check, it blocks a merge the plan declared. Regenerates capability-registry.cjs, which the fragment edit made stale. Neither newly-grown file needs an emitted-drift ack: executor-isolation-dispatch.md sits under workflows/execute-phase/steps/ and execute-wave-pre.md under capabilities/, both outside currentSizes()'s non-recursive scan of gsd-core/workflows/ and agents/. * docs(#3003): document files_deleted where a plan author will actually find it The feature's entire user surface is one plan-frontmatter field, and the canonical reference for that frontmatter -- docs/reference/plan-md.md, the table that documents every other key -- never mentioned it. A field nobody can discover ships as a field nobody uses. Adds the files_deleted row and an example entry in all five locales (en, ja-JP, zh-CN, ko-KR, pt-BR), stating the property that makes the opt-in safe: matching is exact per path after separator normalization, with no globs and no directory prefixes, so a declaration can never authorize more than it literally lists, and omitting the field keeps the guard's original unconditional block. Also corrects two claims in the scope-conformance how-to that this change made false. Its opening paragraph described the recorded declared scope as files_modified alone; declared_deletions is now unioned into that comparison. Its "Renames are not detected specially" bullet asserted the deletions guard blocks any entry whose diff contains a deletion, full stop -- which was the whole point of #3003 and is no longer true. Reworked to say what now decides a rename's fate: declare the old path in files_deleted and both halves become ordinary paths for the advisory check, which is also why the old path needs no separate files_modified entry. Documentation that describes the pre-change behavior of the thing being changed is worse than no documentation, because a reader trusts it. * fix(#3003): close every review finding on the declared-deletions opt-in Two independent isolated reviewers, correctness and security. Neither found a blocker; both found real defects, and the directive treats a finding at any severity as blocking. All of them are fixed here. MAJOR -- the submodule worktree gate could not see a deletion-only plan. per-plan-worktree-gate.md intersected $SUBMODULE_PATHS against $PLAN_FILES alone, while $PLAN_DELETIONS was extracted and then never used. Before files_deleted existed, a path had to appear in files_modified to be planned at all, so the gate saw it; the new field plus the new docs telling authors a deleted path needs no files_modified entry opened a hole where a plan whose only submodule touch is a removal kept worktree isolation on -- the exact case #2772 disabled it for. Both channels now feed the intersection. Note the posture is deliberately the OPPOSITE of the cleanup-wave guard: there the channels stay apart because a deletion AUTHORIZATION must never be inferred; here they merge because a safety fallback must never MISS a touch. MAJOR -- same-wave conflict detection could not see a deletion. The planner's implicit-dependency rule compared files_modified only, so plan A editing src/x.ts and plan B declaring files_deleted: [src/x.ts] scored as conflict-free and ran in parallel: one branch removing what the other is writing, which is the sharpest conflict there is. Overlap is now computed across both channels. MINOR (both reviewers, one root cause) -- the advisory union gave one field two matching rules. declared_deletions was unioned into the scope list handed to planWaveScopeConformance, which reads it with prefix-and-glob semantics. So a field that is exact-match-only at the gate silently became wider at the advisory: ["*.md"], inert at the gate, yielded a null prefix meaning "matches everything" and muted the advisory completely, and ["src"] muted all of src/. The union also activated the advisory on plans that declared no modification scope at all, warning on every modified path. Replaced with subtraction from the findings, gated on files_modified alone. One field, one rule, everywhere. MINOR -- core.quotepath made the feature silently inert for non-ASCII paths. git emits "tests/\303\251.ts" C-escaped and quoted, which never equals the declared plain path, so a correctly declared deletion of tests/é.ts would block forever with nothing pointing at the encoding. Both diffs now pass -c core.quotepath=false. NIT -- flag() consumed a following flag as a value, so --deletions --files x swallowed --files and dropped both. Now treated as a missing declaration, which fails closed. Fixed at both call sites; the helper is duplicated verbatim in cmdWorktreeRecordAgent and cmdWorktreeCreate and leaving one would reintroduce it. TEST -- one test passed for the wrong reason. "a declared deletion is in scope for the advisory" asserted only that warnings omit the deleted path; under a full revert the entry blocks first, warnings come back empty, and the negative assertion passes anyway. It now asserts the entry actually merged, which is the load-bearing half. Four regressions added, one per fix above. Docs corrected rather than extended. The rename bullet in the scope-conformance how-to claimed a rename whose delete side is undeclared never reaches the advisory. Verified false: git's rename detection is on by default, so a pure rename is a single R entry that appears in no --diff-filter=D output and was never gated, before or after #3003. Only a rename that edits enough to fall below the similarity threshold decomposes into add+delete. The pre-existing sentence made the same wrong claim; this restates it correctly instead of sharpening the error. The localized plan-md.md reference edits are reverted: the PR template requires docs content added here to be English, and the translations already lag by three fields, so English-only is the repo's standing posture, not an oversight. Agent-file size caps respected: gsd-planner.md is XL-tier by bytes but carries a separate 49152-LF-CHAR cap asserted by four suites, so its edit is deliberately terse and lands at 49141 with 11 chars of headroom, with the rationale moved to docs/reference/plan-md.md, which has no cap. gsd-plan-checker.md lands at 49107 bytes, 45 under the LARGE cap. Both acks merged into the existing fragments that already name those paths, since two ack sources may never name the same path. * fix(#3003): decode git's path quoting instead of changing the git argv The previous commit's non-ASCII fix turned the remote suite red: 44 failures, 42 of them "unexpected git call: -c core.quotepath=false diff --diff-filter=D --name-only ...". The suite's git mocks match on exact argv, so adding two flags to the deletions diff and the advisory diff invalidated every existing fixture in tests/worktree-safety.test.cjs. Rewriting dozens of fixtures to accommodate one flag would be paying a large Hyrum's-law bill to fix a small defect. Both execGit calls are reverted to their original argv. The C-quoting is now decoded in normalizeScopePath instead, via a new decodeGitQuotedPath helper. That is the better fix on its own merits, not merely the cheaper one: the git argv is untouched so no fixture moves, the decode lands on the ONE normalizer already applied to both sides of the comparison so the declared and reported paths cannot disagree, and it holds regardless of the user's own core.quotepath setting rather than only when we remember to override it. A value not wrapped in a leading AND trailing quote is returned completely untouched, so the plain-ASCII path -- the overwhelmingly common case -- is byte-identical to before. Escapes decode to BYTES collected into a Buffer and UTF-8 decoded only at the end, because \303\251 is two bytes forming one character and decoding them separately yields mojibake. Malformed input never throws: a trailing lone backslash or a short octal escape degrades to the literal character, since one bad path must not take down a cleanup wave. Caught while reviewing the helper: the non-escape branch pushed a UTF-16 code unit rather than UTF-8 bytes. Git always escapes non-ASCII so its own output was fine, but this normalizer runs on the DECLARED side too, and an author may write a quoted path holding a literal é -- pushing 0xE9 alone is invalid UTF-8, so the declaration would decode to a replacement character and silently stop matching. That is precisely the failure this change removes, reintroduced on the other side of the comparison. Now converts whole code points, surrogate pairs intact. The other 2 failures: tests/parallel-dependent-plans.test.cjs pins the exact unbackticked substring "files_modified overlap" in gsd-planner.md, and rewording that comment to "declared-scope overlap" deleted it. The comment is restored verbatim and the files_deleted change rides in the pseudocode and the Rule sentence instead. Recorded in the ack fragment so the next contributor does not rediscover it the same way. Four regression tests cover the decode through the public cleanup-wave seam (the helper is module-private): a declared non-ASCII deletion merges against a C-quoted git report, the symmetric case where the DECLARATION is the quoted form, an undeclared non-ASCII deletion still blocks with the residue naming the decoded path an operator can act on, and a path merely containing a quote is left alone. Plain ASCII was already covered and is not duplicated. * fix(#3003): revert the leading-dash flag guard, the review nit was wrong The remote suite came back with 2 failures, down from 44, and both point at the same thing: tests/worktree-safety.test.cjs:7045 already pins the opposite contract, deliberately. test('a flag-shaped --files value is not re-parsed as a flag', ...) recordAgent(['--files', '--branch']) -> files_modified === ['--branch'] -> branch === 'worktree-agent-a1' ("the real --branch value must be untouched") So consuming the next argv element positionally, whatever its shape, is the tested intent of this parser, not an oversight. The security reviewer's nit claimed --deletions --files x would "swallow --files and drop both". It does not: each flag runs its own indexOf, so --deletions records the literal '--files' while --files independently still resolves to x. And that literal is a path git never reports as deleted, so it authorizes nothing -- already fail-closed with no guard at all. The guard bought no safety and silently changed --files behavior along the way, outside this issue's scope. Reverted at both call sites, which are byte-identical again, along with the test asserting the reverted behavior and the docs sentence describing it. The nit is recorded as REJECTED in the review artifact with the reasoning above, rather than as fixed -- a finding that turns out to be wrong should leave a trace of why, or the next reviewer files it again. docs/CLI-TOOLS.md now states the positional-read behavior plainly instead, so the next person meets it as documented intent rather than rediscovering it through a red suite. * chore(#3003): backfill changeset pr number to 3757 * test(#3003): cover parsePlanDocument's filesDeleted branch to clear the mutation gate CI's Stryker shard for plan-document failed at 73.28 against a break threshold of 75: 170 killed, 62 survived, 232 total. Eight of those survivors are the filesDeleted block this issue added to parsePlanDocument, which shipped with no direct coverage at all -- the field was exercised end to end through the cleanup-wave tests, but the parser itself was never called with a plan that declares it, so every mutant in the block lived. Four tests, each pinned to specific mutants rather than written for coverage percentage: - absent key yields exactly [] -- kills the array-literal seed (["Stryker was here"]) and the `fmDeleted = true` conditional, which would otherwise produce ["true"] - a scalar underscore `files_deleted:` wraps into a one-element array -- kills `fmDeleted = false`, the `&&` logical-operator swap, the `fm[""]` string mutation on the first operand, the emptied if-block, and the ternary's non-array branch - an array-valued hyphenated `files-deleted:` maps element-wise -- kills the `fm[""]` mutation on the SECOND operand (only reachable when the legacy hyphen alias is the one carrying the value) and the ternary's array branch - an empty list yields [] -- boundary case, and a genuinely distinct one from the absent key: [] is truthy in JS so it ENTERS the if, and only Array.isArray's true branch mapping over nothing produces the same [] Threshold arithmetic: 174 of 232 are needed for 75%, and these take it to about 178, so the shard clears with margin rather than landing on the line. Every expected value was confirmed by executing the built parser before being asserted, not inferred from reading the source. --------- Co-authored-by: sim <sim@local>	2026-08-22 13:17:51 -04:00
Tom Boucher	738f42f4fd	feat(#2398): consensus gate for CYCLE_SUMMARY on multi-reviewer runs (#3755) ... * test(#2398): failing-first suite for the CYCLE_SUMMARY consensus gate Binds the gate before it exists, so the suite is RED against next. The load-bearing rows are the two the closed PR #2417 did not have. The B2 regression row asserts a judgment-class lone HIGH counts WITHOUT corroboration when its raiser is unmarked — if anyone re-couples that class to corroboration, more reviewers again produce a weaker gate than one, which is what closed #2417. The parity row asserts every marker literal the gate names is one review-lane-runner actually emits, so the gate cannot key on a signal nothing produces; a mutation row and a seeded fast-check property prove that guard runs its failure branch rather than only reading a correct tree. Also pinned: gate position before Counting rules, the untouched CYCLE_SUMMARY line shape the orchestrator greps, fence balance, the single-reviewer no-op, classification by what a claim asserts rather than by citation presence, the all-marked fail-open, current_actionable staying out of scope, and the leading-marker requirement that stops a review which merely quotes a marker from suppressing its own findings. * feat(#2398): consensus gate for CYCLE_SUMMARY on multi-reviewer runs With review.reviewer_instances running several reviewer identities off one adapter, any single instance's fabricated HIGH could force a full replan cycle on its own. Across ~9 real cycles on two projects each of four instances fabricated at least once, and each was also the most accurate reviewer in some other cycle, so dropping to fewer reviewers trades away real signal. The gate engages only when 2+ reviewers actually ran, and weighs a lone HIGH by what the claim asserts rather than by whether anyone agreed with it. An existence claim -- a symbol, file, flag, commit or ID exists, is absent, or says something specific -- counts only if source-grounding confirms it or another reviewer raised the same concern. A judgment claim -- a design or correctness property -- counts unless that reviewer's own section opens with an evidence-quality discount marker the review lane already stamps ([reviewed-without-source-citations] #3194, [reviewed-without-repo-access] #2176, or a diff-only lane). That split is what resolves B2, the finding that closed PR #2417. B2 showed the approved wording made more reviewers produce a WEAKER gate than one: condition (a) pointed at the source-grounding pass, which verifies every symbol THE PLAN cites and never takes reviewer claims as input, so a genuine architectural HIGH that one reviewer caught and another missed was neither groundable nor corroborated and stopped gating. Judgment-class findings are therefore exempt from corroboration entirely -- reviewers catch materially different classes of issue, and demanding two of them independently raise the same architectural concern suppresses exactly what a multi-reviewer setup exists to surface. Guards on the gate itself: an all-marked cycle disengages it, so a cycle in which nothing was verified can never be counted as converged; the marker must OPEN a reviewer's section, so a review that merely quotes a marker does not suppress its own findings; a suppressed HIGH stays listed and tagged rather than dropped; current_actionable is untouched; and a single-reviewer run is unchanged. No new command, config key, or dependency -- the gate reads signals that already exist. The CYCLE_SUMMARY line shape the orchestrator greps is unchanged; only the integer it computes moves, and only for 2+ reviewers. Known limit, inherited rather than introduced: SOURCE_CITATION_RE checks citation presence, not resolution, which src/review-lane-runner.cts records as a deliberate #3194 scope boundary. A fabricated but plausible file:line still gates. Scope revised and re-approved on the issue before any code was written. * test(#2398): make marker parity behavioral, and stop overclaiming the gate Review found the parity tests were vacuous: they asserted a marker STRING appeared in review-lane-runner.cjs's source text, never requiring the module or calling the stampers, so they would pass even if stampUngroundedReview were broken or never invoked. They now invoke the real exported functions and assert what those functions PRODUCE — that an uncited review gains a leading marker blockquote, that a review carrying a file:line does not, that a self-reported blind review is stamped, and that stamping is idempotent. Removing the source read also removes an incidental no-source-grep evasion via a parameterized path. Review also found the changeset headline false for the class it matters most in. The discount markers detect 'cited nothing' and 'had no repo access'; they cannot detect 'drew a wrong conclusion from a real citation', so a judgment-class finding invented by an evidence-bearing reviewer still counts alone. That is the deliberate side of the tradeoff jags-faith named when closing #2417 — the alternative is requiring corroboration for design findings, which is B2 — but the changeset claimed lone hallucinations no longer force a cycle, full stop. Corrected there, and stated plainly in docs/COMMANDS.md and the design record. Also dropped the reviewer-instances.md entry from the emitted-drift ack: the growth ratchet's currentSizes() scans only gsd-core/workflows/ and agents/ (tests/helpers/emitted-runtime.cjs:916-929), so references/ is outside it and that entry acknowledged a delta the gate cannot see. * chore(#2398): backfill changeset pr number to 3755 --------- Co-authored-by: sim <sim@local>	2026-08-22 10:53:14 -04:00
Tom Boucher	dacae92730	docs(#2845): record the inventory-provenance limits where readers meet them (#3746) ... The limits shipped with #2845 were disclosed only in the PR body, which is read once at merge and then buried. They are properties of what the feature does, so they belong in the documentation. Three surfaces, each at the point a reader forms an expectation: docs/how-to/design-a-ui-phase.md gains a 'What this check is and is not' subsection under the provenance how-to; docs/explanation/security-model.md gains a residual-risk pair matching the section's existing shape; and docs/AGENTS.md notes them where gsd-ui-checker's behavior is described. The substance: a provenance line makes an inventory's origin falsifiable rather than verified, since nothing re-runs the command or compares the count; the rule is agent-applied like the other six dimensions, not a schema check; and 'the checker never runs the recorded command' is an instruction rather than a capability boundary, because the checker holds a Bash grant it genuinely needs for the agent-skills bootstrap and tool grants here are not command-scoped. Co-authored-by: sim <sim@local>	2026-08-21 13:22:50 -04:00
Tom Boucher	4918c62d76	feat(#2845): require provenance for UI-SPEC component inventories (#3745) ... * test(#2845): failing-first suite for UI-SPEC inventory provenance Binds two shared formats before either exists, so the suite is RED against next: the gsd-ui-checker dimension roster (asserted independently on twelve surfaces, eight English and four translated) and the provenance-line grammar the UI-SPEC template emits and Dimension 7 consumes. Every parity assertion is paired with a synthetic mutation case, so the guard's failure branch executes rather than only reading a correct tree: limit-1 (a surface still declaring 6), limit (7), limit+1 (8), a dropped dimension, a label that drifts on one surface only, a non-contiguous roster, a duplicated number, and a surface that stops declaring a count at all. A seeded fast-check property renders the roster under formatting noise (CRLF, padding, interleaved sections) and asserts the parse round-trips and is strictly sensitive to a dropped heading. Assertions are on parsed typed records, never raw substrings. * docs: normalize design-a-ui-phase how-to to American English House style for docs/ is American English (CLAUDE.md). This file carried colour/initialisation/initialise/artefact throughout. Spelling only — no content change; kept separate from the #2845 feature commit so the release-notes classifier and the hotfix cherry-pick filter see it for what it is. * feat(#2845): require provenance for UI-SPEC component inventories A UI-SPEC's component inventory was treated downstream as a closed allowlist while the document recorded nothing about whether the list had been enumerated from the installed design system or recalled from memory. A recalled inventory is indistinguishable from an enumerated one, so an executor complying with the spec builds against a fraction of what the package offers, and every gate stays green because they assert semantics rather than composition. The UI-SPEC template gains a Component Inventory slot carrying one of two provenance lines: the command that enumerated the list, the count it returned, the resolved package@version and the date; or a Could not enumerate record with a real reason. gsd-ui-researcher gains an enumeration ladder and must record the line rather than write the list from recall. gsd-ui-checker gains Dimension 7. An inventory with no provenance line, a count with no command, an empty could-not-enumerate reason, or a line still carrying the template's unfilled placeholders BLOCKs; a partial line, a line placed below its table, or an honest negative record FLAGs; a complete line passes, and so does a spec carrying no inventory at all, which keeps every UI-SPEC predating the dimension validating unchanged. Whatever the verdict, an unsourced inventory is reported as a non-exhaustive list of known-good components rather than a closed allowlist, so the executor is never blocked from a component the spec merely failed to mention. The checker never runs the recorded command. The dimension count moved on all thirteen surfaces that assert it, across five languages. Also corrects the claim in the English, Korean and Portuguese how-tos that this checker applies a scored six-pillar rubric — that rubric belongs to /gsd-ui-review's retroactive audit. * chore(#2845): backfill changeset pr number to 3745 --------- Co-authored-by: sim <sim@local>	2026-08-21 11:59:56 -04:00
Tom Boucher	2b42b28687	fix(#3659): make the worktree base-check trust evidence, not baseRef (#3736) ... * test(#3659): baseref-head suppress must be mode-aware regression rows * fix(#3659): make baseref-head suppress mode-aware and thread isolation mode * fix(#3659): review fixes - stale advice purge, message pins, mode alias * fix(#3659): pick-interceptable emit seam, ack merge, writeSync pin * test(#3659): rewrite set-baseref pin, fix writeSync row stub * chore(#3659): backfill changeset pr number --------- Co-authored-by: sim <sim@local>	2026-08-21 04:58:33 -04:00
Tom Boucher	72819a4616	fix(#3031): opt-in reclaim of GSD hooks orphaned in ~/.kimi (#3731) ... * test(#3031): failing-first coverage for opt-in ~/.kimi legacy reclaim Drives the user-reachable installer surface against a sandbox HOME seeded with the pre-#2755 wreckage: a GSD [[hooks]] block, hooks bundle and CommonJS marker orphaned in ~/.kimi by a --kimi-code install. Covers the reclaim itself plus the four guards the diagnosis identified as negative space: opt-in only (no flag, no deletion), user-authored TOML and hook files preserved, a --kimi install never reclaiming its own root, and the KIMI_SHARE_DIR/KIMI_CODE_HOME collision where both roots resolve to one directory. Adds a fast-check property that stripping the block never destroys user content. Red until --reclaim-kimi-legacy exists. Refs #3031 * fix(#3031): opt-in reclaim of GSD hooks orphaned in ~/.kimi A --kimi-code install older than 1.10.0 wrote its GSD [[hooks]] block, hook bundle and CommonJS marker into Kimi CLI's ~/.kimi. #2755 fixed the destination but could not reclaim what the old bug already wrote: the stale block is byte-identical to a legitimate Kimi CLI one — both runtimes render the same bytes for the same root, since the command paths derive from the hooks root, not the runtime — so no inspection can tell litter from a working install. Cleanup is therefore opt-in. `--reclaim-kimi-legacy` on a --kimi-code install removes GSD's own artifacts from the legacy root; without it nothing is touched, so a dual-product machine keeps Kimi CLI's hooks and #2755's acceptance criterion holds. Extracts the uninstall path's removal sequence into reclaimKimiHooksRoot() and drives both callers through it, so the reclaim removes precisely what a real uninstall removes rather than a hand-copied second implementation. Guards the wrong-runtime case (a --kimi install would delete its own hooks) and the KIMI_SHARE_DIR/KIMI_CODE_HOME collision where both roots resolve to one directory. Also corrects two pre-#2755 leftovers in the same surface that told users to run `--kimi --config-dir ~/.kimi-code` — the form that produces this very defect, since --config-dir moves only the skills root — and adds the missing --kimi-code entry to the installer's own help. Regression coverage folded into tests/kimi-upgrades.test.cjs beside the #2755 cases, per the regression-test-naming lint. Fixes #3031 * fix(#3031): never reclaim ~/.kimi when this run also installs kimi Found by the isolated adversarial review pass and independently while tracing --all ordering, then reproduced. selectRuntimesFromArgs orders 'kimi' before 'kimi-code' in both --all and an explicit --kimi --kimi-code, and installAllRuntimes installs in that order. So --all --reclaim-kimi-legacy installed a fresh, legitimate Kimi CLI hooks block into ~/.kimi and then deleted it moments later from the kimi-code leg — exiting 0 and reporting success while leaving the user with no Kimi CLI hooks at all. The collision guard could not catch it: kimi-code's own root is ~/.kimi-code, a genuinely different directory. The flag asserts "I only use Kimi Code"; installing kimi in the same invocation falsifies that, so the reclaim is skipped with a notice. Also hardens the collision guard itself. It compared path.resolve strings, which returns false for two spellings of ONE directory — measured, not assumed: a symlinked alias and a case variant on a case-insensitive filesystem both compared unequal, so the guard would not have fired and the install would have deleted its own freshly-written hooks. isSameDirectory now compares directories via resolve, then dev+ino identity, then realpath. Regression tests for all three cases; the two alias tests probe the real filesystem and t.skip() where the alias cannot exist. Refs #3031 * docs(#3031): reattach reclaimKimiHooksRoot's JSDoc to its own function Inserting isSameDirectory anchored on the function name, which placed the helper between reclaimKimiHooksRoot's doc block and the function it documents. isSameDirectory ended up with two stacked doc blocks above it and reclaimKimiHooksRoot with none. Refs #3031 * fix(#3031): warn when --reclaim-kimi-legacy cannot apply The flag only acts inside the kimi-code GLOBAL install branch. Passed with any other runtime, or with --local, it was consumed in silence: exit 0, no cleanup, no message. For a cleanup the user explicitly asked for, silence is indistinguishable from "it ran and found nothing". The scope warning is raised at argument-resolution time rather than inside install(). kimi-code declares hostBehaviors.localInstallDeferred, so install() returns early at the deferral check long before the kimi-hooks-toml branch — a guard placed there is unreachable, which is both dead code and a linted drift shape in this repo. Verified reachable by spawning the real installer. Neither case is a hard error: the flag stays composable with --all, where it is legitimately inert for the other seventeen runtimes. Refs #3031 * docs(#3031): document every case where --reclaim-kimi-legacy skips Refs #3031 * fix(#3031): resolve local config dirs from RUNTIME_META alone in the install harness The remote runner surfaced this: the #3031 warning test drives a local kimi-code install and died with "The path argument must be of type string. Received undefined". runMinimalInstall carried a SECOND, hand-maintained local-dir map beside RUNTIME_META, and it had drifted — four runtimes present in RUNTIME_META (hermes, kimi, kimi-code, zcode) were missing from it, so scope:'local' for any of them resolved path.join(root, undefined) and threw a bare TypeError naming neither the runtime nor the map at fault. #3023 had already hit exactly this for pi and fixed it by adding one more entry, which left the divergence itself in place for the next runtime to rediscover. Local scope now reads RUNTIME_META.localDir, the same table the global branch already reads, with the same loud named error the global branch raises. Parity verified for all 14 previously-supported runtimes: every one resolves to a byte-identical configDir. cline keeps its ternary — its local artifacts land at the project root itself, which is a real exception, not a directory name. Guarded in golden-parity-single-source.test.cjs beside the buildParityManifest anti-divergence test, and both arms of that guard were proven able to fail. Refs #3031 * chore(#3031): backfill changeset pr number --------- Co-authored-by: sim <sim@local>	2026-08-21 03:05:58 -04:00
Tom Boucher	9a69a86f42	enhance(#2971): strict planning filter mode for /gsd-pr-branch (#3720) ... * test(#2971): failing-first suite for the pr-branch planning-path filter Binds the not-yet-built planning.pr_strict mode and the corrected filter recipe for /gsd-pr-branch across six layers: pure classification and forbidden-path predicates, real-git fixtures that run the cherry-pick filter loop end to end, config-key registration through the real CLI and both manifests, the executed worktree-materialization claim the issue's triage asked to establish, fast-check properties over arbitrary path sets, and a drift guard over the shipped workflow. Two live defects in today's shipped recipe are pinned as regressions, both reproduced empirically first: `git rm -r --cached` stages a deletion of any .planning/ path the target branch already tracks, so the generated PR removes the base branch's planning files; and the same command leaves the cherry-picked file untracked on disk, so a second commit touching that path aborts the pick with "untracked working tree files would be overwritten" and every remaining commit is silently dropped. The test helper parses the canonical path lists out of gsd-core/workflows/pr-branch.md rather than restating them, so the workflow stays the single source of truth and the suite cannot drift from what ships. Refs #2971 * feat(#2971): strict planning filter mode for /gsd-pr-branch Adds planning.pr_strict — a boolean, default false, that selects what /gsd-pr-branch means by "filtered". Default mode is unchanged: structural planning state survives into the PR branch and the nine transient subdirectories do not. Strict mode drops every .planning/ path, structural files included, and carries a commit over only when it touches at least one file outside .planning/. Strict mode is what makes planning.commit_docs: true safe for a project that versions its planning tree locally but publishes none of it. The alternative posture, commit_docs: false, silently costs parallel executor isolation — a worktree is checked out from a commit, so an untracked or ignored .planning/ is simply absent inside it and the executor has no PLAN.md to read. That claim is now established by an executed fixture rather than inherited. The two path lists are declared once and both projections derived from them, so create_pr_branch and verify can no longer disagree about what the filter promised. verify previously counted every .planning/ path against a documented success criterion of zero while create_pr_branch was specified to preserve five structural files, so a correct run reported itself as failed on every phase that touched STATE.md — which is every phase. It now asserts against the active mode, and names the .planning/ paths default mode deliberately keeps rather than trading a wrong signal for silence. Two verified defects in the same recipe are fixed alongside, because strict mode would have amplified both. `git rm -r --cached` staged a deletion for any .planning/ path the target branch already tracked, so the generated PR removed the base branch's planning files — under strict mode that would have been the entire tree. The same command left the picked file untracked on disk, so a second commit touching that path aborted the cherry-pick with "untracked working tree files would be overwritten" and every remaining commit was silently dropped. Both were reproduced against real git before being fixed. The filter now forces excluded paths back to what the PR branch's HEAD carries, in the index and the working tree; a conflict outside the filter halts instead of being improvised past; a commit left empty by filtering is skipped rather than failing. A clean-working-tree precondition makes the worktree half safe. Closes #2971 * fix(#2971): unwind the checkout on a conflict halt, and test the real recipe Two review findings, both fixed in place. The isolated adversarial pass found that the conflict-outside-the-filter branch exited while leaving the user checked out on the half-built PR branch with cherry-pick state still live — this loop runs in the user's own working directory, so stranding them there is a real cost even though it is not a vulnerability. The branch now aborts the pick, returns to the original branch, removes the partial PR branch, and says so before exiting. The standards pass found the L2 fixtures executed a hand-written mirror of the cherry-pick filter recipe rather than the recipe itself, so a reordering in the workflow would not have been caught — and the order is load-bearing, since restoring a path from HEAD before removing it inverts the filter. The helper now extracts the canonical loop from the shipped workflow and the fixtures execute that verbatim, which also gives the conflict-halt unwind above real coverage. The drift guard additionally pins the two commands' relative order and asserts the workflow carries exactly one canonical loop. Also records the publication gate in the CONTEXT.md glossary next to the commit gate it is distinct from. Refs #2971 * fix(#2971): make the conflict-halt unwind actually unwind, and use the colon slash form The remote matrix caught two defects in the previous commit. The halt path claimed to restore the original branch but did not. `git cherry-pick --abort` does not apply to a single `--no-commit` pick with no sequencer file, and the fallback left the unmerged index in place, which makes `git checkout` refuse — a failure the `2>/dev/null || true` then swallowed, so the user was told they had been restored while still sitting on the half-built PR branch. The unwind now drops sequencer state, hard-resets the disposable PR branch to clear the unmerged index, and only claims a restore when the checkout actually succeeded; when it does not, it says where the user is and gives them the two commands to finish it by hand. Verified against real git: exit 1, the conflict named, HEAD back on the original branch, the partial branch gone, a clean tree and no CHERRY_PICK_HEAD. Two runtime-loaded source artifacts used the retired `/gsd-<cmd>` hyphen form, which names a command no runtime registers. The canonical authoring token for workflows and references is `/gsd:<cmd>`; docs keep the hyphen form, so the documentation added in this branch is unaffected. The comment in src/config.cts moves to the colon form too, since it propagates into the generated lib. Refs #2971 * docs(#2971): backfill PR number into the changeset fragments (#3720) --------- Co-authored-by: sim <sim@local>	2026-08-20 15:07:40 -04:00