161 Commits

Author	SHA1	Message	Date
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	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	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	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	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
Tom Boucher	14679b866b	enhance(#2856): add default-off live-DOM UAT capability (#3716) ... * test(#2856): add failing-first suite for the live-dom-uat capability Binds the approved triage shape before any of it exists: - containment — the execute:wave:post hook must not render unless workflow.live_dom_uat is true AND the capability resolves active (fail-closed on a missing state entry, and on a non-boolean value) - criterion 4 — agents/gsd-executor.md carries no browser MCP family; asserted as an absence, which is the only way it is observable - Hyrum guard — the pre-existing mcp__playwright__* branch must stay outside the key-gated block, or upgrading silently removes working automated UI verification for every current Playwright-MCP user - parity — the browser glob list now lives in two surfaces (agent frontmatter + workflow detection block); the assertion fails if either gains or loses a family without the other Red by construction: the capability, agent and workflow block do not exist yet. Verified on the remote runner. Refs #2856 * enhance(#2856): add default-off live-DOM UAT capability A phase whose acceptance criteria needed a live DOM could not be finished by the agent that executed it: gsd-executor carries no browser tools, so it correctly returned checkpoint:human-action even though the work was not human-only, just tool-less. Every such phase degraded to "executed, then finished by hand in the orchestrator", and autonomous: false could not distinguish "a human must judge this" from "the executor lacks the tool". Implements the shape approved at triage, not the one reported. The executor's tools: line is NOT widened, in any configuration: for a first-party agent the static list is the only control that exists (ADR-1244 D2, ADR-857 D4, no per-dispatch override). Instead one default-off capability owns the key, the agent, and the step: - capabilities/live-dom-uat/ — activationKey workflow.live_dom_uat (boolean, default false), one additive step at execute:wave:post (onError: skip, gates: []), so it can never halt a wave - agents/gsd-dom-verifier.md — the only GSD agent carrying browser MCP globs, in its own tools: line, with no Bash - verify-work automated_ui_verification — a gsd:live-dom-families block naming both new families AND the key; presence alone never activates Two independent fail-closed gates: isCapabilityActive renders a hook only on state.active === true, plus the step's own `when`. The pre-existing mcp__playwright__* branch keeps the gating it already had and stays outside the new block. Pulling it behind a default-off key would have silently removed working automated UI verification from every current Playwright-MCP user on upgrade. Also closes a host gap this surfaced: execute:wave:post dispatched only contribution + gate, so ANY registered step was declared and silently never run — exactly the single-kind hand-roll loop-hook-dispatch.md names. Step 5.75 now dispatches every kind == "step". The browser-profile lock is tolerated, not coordinated: --isolated is a flag on the operator's own MCP-server registration that GSD neither launches nor parameterizes, so the verifier reports could_not_look / profile_locked, names the flag, and stops. DOM-VERIFY.md keeps could_not_look and nothing_to_report distinct behind a closed reason enum — collapsing them is the ambiguous-run-notes defect reported. Verified on the remote runner. Closes #2856 * fix(#2856): apply review findings from the orthogonal passes Correctness pass (blocker): - delete detectionBlockIsCrlfSafe. It was pass-always: it read the file, replaced LF with CRLF, then indexOf'd marker strings that contain no newline, so the replacement could not change the result and the assertion could never fail for the reason it stated. There is no real CRLF risk on this surface either — the gsd:live-dom-families block has no parser, only human and agent readers. Deleted rather than replaced, per the repo's pass-always-test rule. Isolated security pass (two minors, both real): - execute-phase.md step 5.75: this change is what first activates kind == "step" dispatch at execute:wave:post, which newly opens the ref.command shell path at that loop point. Our own step uses ref.agent and never touches it, but the door is now open, so the step-dispatch line carries the same in-context validate-before-shell warning the sibling gate-dispatch line directly below it already carries. - gsd-dom-verifier: quoted page text in DOM-VERIFY.md is attacker influenced. Require it wrapped in inline code or a fence, kept short, and never left reading as a directive to the next reader. Verified on the remote runner. Refs #2856 * fix(#2856): settle the new-agent roster ripple Checkpoint 2 returned 28 failures, none in the new suite — all of them the guards that exist to make adding an agent a deliberate act. Each is a real boundary that had to move: - docs/AGENTS.md: Tools row must copy the frontmatter verbatim (#2526), so the browser globs lose their backticks; primary-agent counts 21->22, roster 33/34->34/35, Verifiers category 1->2 - docs/INVENTORY.md: roster completeness requires every agents/gsd-*.md to be classified exactly once - gsd-dom-verifier: add the anti-heredoc instruction and the commented hooks: frontmatter pattern both agent gates require - gsd-core/bin/shared/model-catalog.json: every shipped agent needs a profile entry (#3229) - copilot-install / kilo-upgrades / qwen-upgrades: expected agent list and the 34->35 roster boundary - execute-wave-post-gate-pipeline-e2e: execute:wave:post legitimately carries one step now. Asserted as an exact shape — one step, capId live-dom-uat, ref.agent gsd-dom-verifier, onError skip — so it stays a real guard against accidental change rather than being relaxed Two findings worth naming: mcp-tool-inheritance (#2526) rejected the agent for documenting mcp__playwright__* while its tools: line withholds it — a dead instruction that invites the agent to claim a path it cannot take. The prose now names the Playwright MCP family without the dispatchable token, in both the agent and the capability fragment. runtime-launcher-parity rejected the new gsd_run call: each fenced block is its own shell, so a workflow step file invoking gsd_run needs its own canonical preamble. Propagated with scripts/sync-runtime-launcher.cjs. That script also normalizes explore.md, which is unrelated pre-existing drift the parity check tolerates, so it is reverted to keep this diff scoped. The emitted-drift ack supersedes the spent #3370 entry for execute-phase.md — it is merged into next, so its ripple is absorbed at the base and it can no longer clear anything. That is the same supersede the #3370 entry itself performed on the spent #3324 fragment. Its unrelated execute-plan.md entry is untouched. Verified on the remote runner. Refs #2856 * fix(#2856): drop the stale emitted-drift ack entry The automated-ui-verification.md entry was written speculatively rather than from a reported growth, and the check names that precisely: an ack "written or reworded in THIS diff, but nothing here needed it, so it explains nothing". The growth tier keys on the bare filename as it appears under gsd-core/workflows/ or agents/. automated-ui-verification.md is nested under verify-work/steps/, so it was never in the tracked set — only execute-phase.md was ever reported, both before and after the launcher preamble landed. Only ack what the check actually reports. Verified on the remote runner. Refs #2856 * chore(#2856): backfill changeset pr number pr:0 -> 3716. The placeholder fails both changeset-lint (fail_invalid_fragment) and docs-lint (fail_malformed_fragment) by design and can only be resolved once the PR number exists. Both now report ok against GITHUB_BASE_REF=next. Refs #2856 --------- Co-authored-by: sim <sim@local>	2026-08-20 15:07:21 -04:00
Tom Boucher	8df5cb36c2	enhance(#2951): refuse [VERIFIED] for a compatibility claim resting on absent metadata (#3718) ... * test(#2951): pin the absent-evidence provenance contract (failing first) 17 tests / 22 anchors on the deployed agent text. Measured against the parent commit: 20 anchors fail, 2 pass. The two that pass are the sibling-integrity guards on the package-name and in-repo-value rules -- green before and after is their intended signature. Refs #2951 * enhance(#2951): refuse [VERIFIED] for a compatibility claim resting on absent metadata A claim of the form "X does not support Y" drawn from MISSING metadata -- no python_requires, no engines field, no per-version classifier, no changelog entry, no matching support-matrix row -- no longer earns [VERIFIED] however authoritative the source consulted. An absence is silence about every value, so the same evidence would "prove" both the version being ruled out and the version being standardized on. The only route from an absence to [VERIFIED] is a positive falsification attempt with its failing output pasted; everything short of that is [ASSUMED], which the file already routes to "needs user confirmation before becoming a locked decision". Third member of the family beside the package-name and in-repo-value provenance rules, mirroring PR #2768's shape. A present declared constraint and an affirmatively documented incompatibility are untouched. Closes #2951 * fix(#2951): close the allow-list ambiguity and the mutation gap review found Findings from the isolated adversarial pass and the two-axis review, all fixed: MAJOR (x2, one root cause) -- the absence clause and the present-constraint carve-out gave opposite verdicts on the same evidence for the commonest real case: a classifier list enumerating :: 3.9 through :: 3.13 with no :: 3.14. A researcher could read the enumerated list as a "declared" positive constraint and re-earn [VERIFIED], which is also the evasion vector. The rule now states the decision procedure -- does the declaration bound EVERY value or only the ones it names -- and closes the positive-reframing restatement explicitly. New contract test pins all four clauses. MAJOR -- 'licenses a positive falsification attempt as the route to [VERIFIED]' asserted two independent substrings and never that the route lands on [VERIFIED]. A mutant swapping the tag for [CITED] or [ASSUMED] inverted the rule and survived all 17 tests. Now pinned as one joined sentence. MINOR -- the attributable-failure test regex-matched illustrative examples ("a missing certificate, a wrong host"), so a copy-edit would break it for no reason; relaxed to the substantive clause. The no-paraphrase guard counted only the heading, missing the drift mode in its own name; it now also pins the core proposition to one occurrence, and the test name matches what it checks. An off-by-one in the new allow-list regex bound (141 actual vs 140) is fixed. MINOR -- docs/AGENTS.md listed four of the five governed absence forms while the agent prose, docs/COMMANDS.md and the changeset listed five; three copies disagreeing on list membership is the drift this repo treats as a defect. SCOPE -- removed docs/how-to/verify-a-dependency-compatibility-claim.md and its docs/README.md index line. Both reviewers flagged them as a seventh and eighth surface beyond the six the requester capped, and CONTRIBUTING's "Agent or skill change" row requires only docs/AGENTS.md. The actionable four-case guidance is retained in docs/COMMANDS.md, which is inside the approved scope. Ack byte figures corrected for the final size: 44250 -> 46602 (+2352), 2550 bytes headroom under the LARGE cap of 49152. Refs #2951 * docs(#2951): restore the how-to the phase gate requires Reverses the removal in 6404b43d3. Both /code-review axes had flagged docs/how-to/verify-a-dependency-compatibility-claim.md and its docs/README.md index line as a seventh and eighth surface beyond the six the requester capped, and CONTRIBUTING.md's required-docs row for an "Agent or skill change" names only docs/AGENTS.md, so they were dropped. gsd-phase-gate.cjs then denied gh pr create: it refuses when the recorded enablement sequence has more than one step and the how-to quadrant is empty. The sequence here is genuinely four steps -- run plan-phase, read the [ASSUMED] claim, probe or cite or accept it unlocked, then answer discuss-phase's checkpoint -- and the last step lands on a different capability's surface, so a reference table cannot carry it. Compressing the sequence to one step to unlock howToSkipReason would be gaming the gate, which is the same Goodhart failure this whole change exists to close. A machine-enforced repo gate outranks two reviewers' scope preference and my own reading, so the page is restored and the PR body discloses the two extra surfaces instead of hiding them. Reverting is a one-file change if a maintainer prefers the tighter scope. Refs #2951 * chore(#2951): backfill the changeset PR number pr: 0 -> 3718 now that the real PR exists. The placeholder fails scripts/changeset/lint.cjs with fail_invalid_fragment, which also blocks lint-docs-required from consuming the fragment. Refs #2951 --------- Co-authored-by: sim <sim@local>	2026-08-20 14:35:14 -04:00
Tom Boucher	8da2dd3ad2	feat(#2790): add read-only planning.inspect schema-v1 snapshot query (#3708) ... * feat(#2790): add read-only planning.inspect schema-v1 snapshot query Adds a read-only query emitting a schema-versioned JSON projection of .planning/ so downstream harness UIs can consume planning state without parsing GSD's Markdown a second time. Composed strictly from the ADR-3180 section 7 owners plus parsePlanDocument, parseRequirements and parseUatItems; markdown structure is read through the Markdown Sectionizer and Markdown Table Model seams. It declares its own flat external schema rather than serializing PlanningSnapshot, which is the diagnostic-rule subject and still growing. Extracts plan-document parsing out of cmdPhasePlanIndex into a shared leaf module so phase.plan-index and planning.inspect cannot drift, including the plan-id derivation both surfaces report. Also fixes parseRequirements dropping the separator delimiter used by the shipped requirements template, surfaced while wiring the requirement rows. * fix(#2790): close spec gaps and a raw-text test assertion found in review Review findings from the standards, spec and security passes: - phases[] rows carry goal and dependencies, the two per-phase elements the issue Summary names that had no corresponding field. Goal is bounded to the section's leading prose so the Depends-on line, the Plans checklist and the wave annotations are not duplicated into it. - requirement rows carry their own diagnostic codes, so a consumer no longer has to string-parse the global diagnostics subject to correlate. - roadmap_acceptance.checkbox is looked up through the phase-id key owners. It was compared raw against the on-disk directory name, so it read null for every real-world slugged phase directory and the evidence channel was inert. - the hostile-input test asserts the structured payload instead of matching the raw stdout string. The absence proof over raw stdout is kept deliberately. * fix(#2790): register planning in the runtime usage list and repair fixtures Remote runner reported 9 failures on 9b3f9aa. Two root causes, both fixed: - gsd-tools.cjs registered the planning family in HOST_COMMAND_ROUTERS but never added it to TOP_LEVEL_USAGE's Commands list. Those are two surfaces a parity test guards, and the top-of-file block comment is not the runtime help string. A real wiring gap that every local gate and three review passes missed. - the new suite's fixtures could not produce a resolvable phase set. STATE.md frontmatter omitted the milestone field, which ADR-3180 7.2 rule 1 makes the primary milestone selector, so the phase set scoped unscoped and every percentage was correctly withheld. Separately declarePhase returned a path without creating the directory, so a phase declared but never written to left phases empty. Both reproduced against the built module before fixing. No assertion was weakened. The withholding path is still exercised and still returns null when the roadmap is absent. * chore(#2790): backfill changeset pr number * test(#2790): cover every enumerated matrix row and contain a symlink escape Reverses a silent deferral. An earlier revision left 23 of the 78 enumerated matrix rows unimplemented and 7 more as one-off manual checks, with a paragraph in the artifact and the PR body describing the gap. CLAUDE.md is explicit that such a note is not a fix and is not surfacing. The rows are implemented instead and the manual-evidence bucket is gone: 49 test cases become 88, covering all 78. Writing the symlink row proved a real leak: a *-PLAN.md symlinked outside .planning/ had its content emitted into the payload, confirmed via a direct call and the spawned CLI. readDocument now resolves target and planning root with realpathSync and rejects an escape, returning the ordinary unreadable-document shape. Tested both ways, because a containment check that over-rejects is its own defect: an escaping symlink leaks nothing and degrades that plan alone, while a legitimately relocated .planning/ symlink stays fully readable. The three new modules are registered in the mutation COVERED registry, which had been reporting has_work false and skipping the Stryker gate entirely. Provisional non-binding floors so the shards run and report; raised to the measured value before merge, since the registry forbids calibrating from a local run. * fix(#2790): satisfy the mutation ratchet contract and scope the 1MB test Remote runner reported 16 failures on 8c451ed. Two causes. The COVERED registry has a paired contract the earlier commit violated: every module needs a matching RATCHET_BASELINE entry, and minScore must be between 50 and 100 with minScore === baseline. The provisional floor of 1 was illegal on both counts. All three modules now sit at 50 — the registry's own enforced minimum — with matching baselines. The score cannot be measured locally: the shard runs node --test, which this repo hard-blocks, so CI is the only source. Floors are raised to the measured value once this PR's shards report; a shard below 50 means the tests need strengthening, since the floor cannot go lower. The 1MB test was measuring the test harness rather than the product. The command handles the oversized payload correctly by spilling to a tmpfile and resolving it back, but the resolved stdout then exceeds runGsdTools' maxBuffer and the helper reports ENOBUFS. It now uses --pick so stdout stays one byte while the full 1MB document is still read and parsed end to end. * fix(#2790): wire containment across every document read this command drives An isolated security review of the containment control found the boundary logic sound but not comprehensively wired: two content reads reached the filesystem without it. An escaped phase DIRECTORY could enumerate external filenames into the file fields and diagnostic subjects. Both enumeration sites now containment-check the directory before reading. Worth recording that the leak was already prevented one layer earlier than the review claimed: Dirent#isDirectory() reports false for a directory symlink, so such a directory never becomes a phase row at all. The guard is defense-in-depth for a direct caller and for platforms where a reparse point reports as a directory. A *-VERIFICATION.md symlinked outside the root leaked one frontmatter value verbatim, because readVerificationStatus does its own read and copies an unrecognized status into the payload's next_action. Closed from the consumer side through that function's existing fs injection seam, so src/verification.cts keeps its signature and its other callers are untouched. The reviewer additionally rated a forged status: passed as an integrity bypass. It is not: anyone able to plant the symlink can plant a real VERIFICATION.md saying the same thing. The incremental risk is confidentiality, which is what these fixes close. src/plan-scan.cts is deliberately unchanged: isPlanSuperseded reads symlink-followed content but yields only a derived boolean, no document text. * test(#2790): give the mutation shards an in-process surface Two Stryker shards were CANCELLED at the 15-minute cap, not failed on score. CI log: 640 mutants instrumented, and the dry run reported 'Ran 1 tests in 20 seconds' because the shards pointed at the integration suite, where nearly every case spawns a gsd-tools subprocess and Stryker's command runner treats the whole test-runner invocation as a single test. 640 x 20s cannot finish in 15 minutes; at the kill it was 27/640 with an ETA over an hour. Every other COVERED module points at a property or unit file, and the workflow's own paths filter lists exactly those two patterns. In-process is the intended mutation surface; the shards were pointed at the wrong shape of test. Adds tests/planning-inspect.unit.test.cjs — 39 cases in 10 describes that spawn nothing and call the built modules directly. plan-document and the router need no filesystem at all, one being a pure content-to-object parser and the other taking an injected mock. The three shards now point here. The 91-case integration suite is untouched and still runs in the normal test job. * chore(#2790): ratchet mutation floors to the measured CI scores CI run 32392791843 measured all three shards, which is the only source the registry accepts — local runs count timeouts as kills and inflate badly. planning-command-router 95.65 -> floor 94 plan-document 76.58 -> floor 75 planning-inspect 57.03 -> floor 56 Applied the registry's own rule, floor(score) - 1, and updated RATCHET_BASELINE to match, since the ratchet test enforces equality. planning-inspect sits well below the file's target of 80 and is the obvious ratchet candidate as its tests improve. planning-command-router already exceeds the target. The placeholder comment about floors pending measurement is removed rather than left standing as a false statement. --------- Co-authored-by: sim <sim@local>	2026-08-20 13:42:43 -04:00
Tom Boucher	77fa08f1e8	fix(#2773): feed the spec-phase edge probe English-translated requirement text (#3713) ... * test(#2773): failing-first contract and premise tests for translated edge-probe input Locks the Step 5.5 contract that a response_language project must feed the edge probe an English translation of each requirement's text, and binds that advice to measured engine behavior: the same requirement classifies to zero shapes in Portuguese and to collection/adjacency/empty/ordering in English. Also pins the honest limit — the issue's own repro sentence classifies to [] in English too, so translation is necessary but not sufficient and the authored shapes override is the documented fallback. Red before the doc change; the assertions are all false today. Refs #2773 * fix(#2773): feed the spec-phase edge probe English-translated requirement text The shape cues in src/edge-probe.cts are English word-boundary regexes, so a project running with response_language set wrote its SPEC requirements into the Step 5.5 $REQS_JSON heredoc in that language, matched no cue, classified to zero shapes, and landed every row in the unclassified sentinel (#1110). The taxonomy contributed nothing and --auto left it all unresolved — the probe was a silent no-op for exactly the spec type it exists to harden. Step 5.5 now states that the $REQS_JSON payload is engine input rather than user-facing output, so the response_language rule does not govern it: each requirement's text carries a faithful English translation, the SPEC keeps its original language, and requirement ids are never translated or renumbered. The instruction sits before the heredoc on purpose — the downstream APPLICABLE=0 warning fires only when every requirement is unclassified, so a partly-classified non-English spec would otherwise slip through with no signal at all. Measured against the compiled engine: the same requirement returns [] in Portuguese and collection -> adjacency/empty/ordering in English. Also measured: the issue's own repro sentence returns [] in English too, so translation is necessary but not sufficient — the instruction therefore points at the authored shapes override for prose carrying no cue in any language rather than promising that translation restores classification. Doc scope only, per the triage disposition on the issue. The compiled engine is untouched; the lang-hint / per-language cue-set fix is a separate follow-up. Closes #2773 * fix(#2773): clean up the edge-probe temp file on the placeholder-guard exit path Surfaced by the isolated security review of this branch. Between the mktemp and the unconditional cleanup, Step 5.5 has two sibling guards that disagreed about their own invariant: the engine-failure guard runs rm -f "$REQS_JSON" before exiting, while the empty/placeholder guard directly above it exited without one. A spec run that tripped the placeholder check therefore stranded a temp file holding the SPEC's requirement text in TMPDIR, once per failed run. The added contract test walks the region between the mktemp and the unconditional cleanup and asserts no exit path leaves the file behind, so the two guards can no longer drift apart. Proven to bind: run against the pre-fix file the walker reports the leaking exit; against the fixed file it reports none. Refs #2773 * docs(#2773): record the edge probe's English-cue input constraint in the predicate store The co-change gate flagged CONTEXT.md (13 co-changes with spec-phase.md) and docs/CONFIGURATION.md (11) as candidate-missing-updates, and both were real gaps rather than incidental coupling. CONTEXT.md's EdgeCompletenessProbeModule entry documents the input contract for classifyShape but did not record that SHAPE_CUES are English word-boundary patterns — so the predicate store implied text was language-agnostic, which is what a future agent reads before touching this seam. docs/CONFIGURATION.md's response_language row is what a non-English project reads when it turns the setting on; it now names the one deliberate exception and links to the FEATURES.md explanation, so the interaction is discoverable from the config key rather than only from the workflow. CONTEXT-INDEX.json regenerated via gen-context-index.cjs --write. The drift-ack fragment is updated for the final byte range and now also records the placeholder-guard cleanup fix folded into the same block. Refs #2773 * fix(#2773): append the growth rationale to the existing spec-phase.md ack entry The remote runner caught this: emitted-attribution.test.cjs pins the 0000-legacy-migration.json spec-phase.md entry permanently (the #2914 migration regression test asserts the exact '31987 -> 31997' delta text survives), so removing it to avoid a duplicate-key collision with a new fragment broke that test instead of satisfying the ratchet. The entry is an accreting log, not a single-use slot — #2733, #3132 and #3102 were each appended to the same reason string by later PRs, which is how a shared growth key coexists with the rule that two ack sources may never name the same path. This appends the #2773 rationale the same way and drops the separate fragment, whose spec-phase.md key was the collision. Verified locally by reproducing both affected tests against the real fragment before re-dispatching: the pinned delta survives, grown[0].acked is true, staleAcks is empty, and all 35 entries still read as spent. Refs #2773 * docs(#2773): add a how-to for probing edges in a non-English project The phase gate's enablementSequence check caught a wrong call of mine. I had recorded that no how-to was owed because the user takes zero extra steps — the workflow translates the probe input itself. Written out, though, the sequence from off to value is two steps and step 1 depends on response_language, a setting owned by a different capability than the edge probe, which is exactly the condition the how-to test names. There is also real task content a reference table cannot carry: the three-way split between a few unclassified rows (the classifier's recall gap), every row unclassified (the probe could not read the spec at all), and the silent partly-classified case where the APPLICABLE=0 warning never fires. That last one is what a user would otherwise misread as a clean bill of health. Shaped after the resolve-edge-coverage-findings / resolve-unreachable-guard siblings and indexed from docs/README.md next to its closest relative. Refs #2773 * chore(#2773): backfill the changeset PR number pr:0 placeholder replaced with the real PR number now that #3713 exists. Refs #2773 --------- Co-authored-by: sim <sim@local>	2026-08-20 13:36:00 -04:00
Tom Boucher	adb46cdd85	feat(#2734): surface STATE.md commit-age on the statusline (#3700) ... * test(#2734): failing-first suite for the statusline STATE.md freshness marker Binds the contract before any hook change exists: a `state ~N commits back` segment gated on the state_head stamp landed by #2622, firing at the same advisory threshold /gsd-health's W024 uses rather than at > 0. Covers all five acceptance criteria — threshold parity (19/20/21 boundaries), both renderers including formatGsdStateCompact, an exact spawn-count assertion, repo-pinning and sub_repos degradation, and behavioral parity against readStateHeadFreshness rather than a source-grep of the two fence copies. 52 example-based tests plus 5 seeded fast-check properties. Red now by design. * feat(#2734): surface STATE.md commit-age on the statusline Adds an opt-in `state ~N commits back` marker to the GSD-state segment, consuming the `state_head` stamp and freshness contract landed by #2622. A solo developer returning to a project reads "Phase 4, executing" in STATE.md and acts on it, without noticing the codebase moved 40 commits since that line was written. /gsd-health reports it as W024, but only if you think to run it; the statusline is the surface you see without asking. Fires at STATE_HEAD_ADVISORY_COMMITS (20), the same threshold W024 uses, not at > 0: with commit_docs:true the commit carrying a STATE.md sync advances HEAD by one, so > 0 would alarm permanently on a fresh project. Costs exactly one bounded git subprocess per render and none when disabled. `rev-list --left-right --count` answers ancestry and distance together, and repo pinning is a filesystem check mirroring projectOwnsItsRepo rather than a --show-toplevel compare, which is unreliable on macOS /private/var and Windows 8.3 paths. Every unresolvable input degrades to the tri-state unknown -- the marker is absent, never a "fresh" claim the project cannot substantiate: a malformed stamp, a root that does not own its .git, a sub_repos workspace, history rewound past the stamp, or git being unavailable. Also collapses statusline config resolution onto one resolveStatuslineOptions() seam. runStatusline() and renderStatusline() duplicated it byte-for-byte; one copy is what keeps a newly-added key from reaching only one of them. * test(#2734): route the e2e spawn through the process seam and fix fixture leaks Review findings from the two orthogonal passes: - `bothEntryPointsResolveOptionsIdentically` spawned a child and substring-matched its stdout to test a pure function. It now calls resolveStatuslineOptions() directly — no subprocess, no text matching. - `skipsFreshnessWorkWhenTodoTaskActive` genuinely needs a child (the !task gate lives in runStatusline, which reads stdin), so it now spawns through tests/helpers/process-seam.cjs and proves the negative with a filesystem fact: the git shim appends to a marker file on every invocation, and the assertion is that the marker never appears. Stronger than asserting text is missing, and it drops the last stdout substring match in the block. - Every fixture-creating test now registers `t.after(() => cleanup(dir))` instead of a trailing cleanup(dir), which leaked the temp repo on assertion failure. derivationAgreesWithStateModule reassigns `dir` across five fixtures, so it binds each directory at scheduling time rather than cleaning only the last. Also corrects markerCoexistsWithMilestoneComplete, which asserted the wrong expectation rather than finding a code defect: `percent` drives the progress bar too, so the milestone segment reads "v1.9 [##########] 100%". The marker appends after it, which is what the test exists to prove. CONTEXT.md's opt-in statusline key list was missing statusline.show_git as well as the new key; both are now enumerated. * docs(#2734): backfill changeset PR number (#3700) --------- Co-authored-by: sim <sim@local>	2026-08-20 00:35:01 -04:00
Tom Boucher	2fca0e17e4	enhance(#2554): resolve code review depth from path-scoped override rules (#3695) ... * test(#2554): failing-first suite for path-scoped code review depth overrides Binds the not-yet-built code-review-depth module: segment-aware path-prefix matching of a changed-file set against ordered {paths,depth} rules, resolution order flag > strongest matching rule > global > standard, typed validation errors, and the large-scope downgrade boundary. Also proves behaviorally that workflow.code_review_depth_overrides is not yet a registered config key. Refs #2554 * feat(#2554): resolve code review depth from path-scoped override rules Adds workflow.code_review_depth_overrides — an ordered array of {paths, depth} rules matched against a review's changed-file set by segment-aware path-prefix comparison. Resolution order is --depth= flag, then the strongest matching rule, then workflow.code_review_depth, then standard; a matching rule replaces the global rather than being max'd with it, so quick and standard rules stay meaningful. Glob metacharacters are a hard configuration error rather than sugar for a prefix, and malformed rules halt the review instead of degrading to standard. The resolver is pure and reports its own provenance, so the workflow can print the resolved depth and the rule that matched. The pre-existing >50-file deep-to-standard downgrade moves into the module and now names the rule it overrode. The key is registered centrally rather than as a capability config slice: the federated slice channel admits only boolean/string/number/enum, so an array slice would be dropped as malformed. Closes #2554 * test(#2554): correct depth-provenance assertions and pin out-of-repo paths Two corrections to the failing-first suite. The source assertion for a non-matching rule with no global configured expected 'config'; with no global set the depth comes from the default, and a companion assertion tolerated either value, so both passed against an implementation that derived provenance from whether any rules existed rather than from where the depth came from. The out-of-repo absolute-path case used a home-directory path that matched neither implementation, so it never exercised the defect it named. It now pins the discriminating cases: an absolute path outside the repo root must not match a repo-relative rule, and one under the root must. * docs(#2554): document path-scoped code review depth overrides Reference rows for workflow.code_review_depth_overrides in the configuration, features and commands references plus the locale copies that carry those tables, and in the planning-config reference. Explanation of why escalation is whole-review rather than per-file and why v1 is prefix-only. New how-to for scoping review depth by path, carrying the configuration-error reason table and the distinction between nothing to report and could not look. CONTEXT.md glossary entry and the INVENTORY row for the new CLI module. ja-JP and ko-KR CONFIGURATION.md carry no code_review keys at all, and ko-KR and pt-BR FEATURES.md carry no code-review config table, so those files are deliberately untouched. * fix(#2554): make the depth-misconfiguration halt executable and reject control chars Three review findings, all in this change. The misconfiguration halt was prose rather than shell: the error-printing fence was followed by an unconditional extraction fence, so an ok:false result threw and left the depth empty instead of stopping the review. Prose is not a guard — the two fences are now one block with a real conditional, and anything that is not the literal string true fails closed. An interior control character in a rule path survived validation and reached the provenance string and the summary box; rule paths now reject control characters via a new PATH_CONTROL_CHAR reason, after the glob check so precedence is unchanged. That in turn makes the field record safe to delimit, so the seven node invocations that each re-parsed the same result to read one field collapse to one. Also corrects the glossary entry's illustrative paths, which the glossary-ref check read as real repository references. * fix(#2554): use the fast-check v4 string API and acknowledge workflow growth Two failures from the remote matrix on d3111f45, both this branch's. The property block built its segment arbitrary with fc.stringOf, removed in fast-check v4. Because the arbitrary is constructed in the describe body, the throw took out all four property tests rather than one — they had never executed. Rewritten to fc.string({unit, ...}), the form this repo already uses in emitted-attribution.test.cjs. Every other fast-check helper in the file was audited against the installed module. The emitted-attribution growth arm needed an acknowledgment for code-review.md, which grew 5376 bytes. The pre-existing 3503 fragment keying the same file is spent — its ripple was absorbed when #3503 merged, and the base file is exactly the 34435-byte baseline this growth is measured against — so it cannot clear anything, while the ack lint hard-fails on a duplicate key across two sources. Removed it in favor of the new fragment, which is exactly how #3503 itself replaced the spent 3191 fragment. * docs(#2554): backfill changeset PR number --------- Co-authored-by: sim <sim@local>	2026-08-19 22:45:35 -04:00
Tom Boucher	79781e68eb	enhance(#2401): ground verify-command paths and inherit prior-phase commands (#3678) ... * feat(#2401): ground <automated> verify-command paths and inherit prior-phase commands Adds a deterministic resolvability probe over each PLAN.md <automated> verify command and surfaces the nearest prior phase's proven commands to the planner at every context window. - src/verify-command-grounding.cts: recognizer (not a shell interpreter) that grounds a leading cd <literal> chain and npm --prefix <literal>, and reports unresolvable rather than guessing. Never executes command text. - gsd-tools check verify-command-paths <N>: per-phase probe, wired into plan-phase.md before the plan-check pass. - init.plan-phase gains prior_verify_commands, ungated by context_window. - gsd-plan-checker: new Verify Command Path Resolvability dimension that reports the failing target and never prescribes a replacement. Also fixes first-match-wins prefix bucketing in scripts/lint-test-file-count.cjs (readdir order is not stable across platforms, so a module whose name extends another's with a hyphen bucketed differently on Linux than on macOS). Closes #2401 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#2401): ground the canonical --prefix form, quoted paths, and absolute cd resets Independent review found three defects in the recognizer: - npm --prefix DIR run SCRIPT never reached the script-existence check, because the pattern required npm and run to be adjacent. That is the form the docs tell planners to prefer, so script_missing never fired for it. The prefix flag and its value are now stripped before matching. - --prefix captured with \S+, so a quoted path containing a space was truncated to a stray opening quote and reported as a missing directory - a false blocker, worse than the bug this feature fixes. The capture is now quote-aware. - A chained cd whose later segment was absolute concatenated instead of resetting, producing a nonsense path and another false blocker. The fold now resets on an absolute segment. Also replaces the bespoke phase-directory regex with the canonical phase-id helpers. Real phase directories are NN-slug, not phase-N-slug, so the prior-command harvest matched nothing outside its own fixtures and the planner-inheritance half of this feature was dead code. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * refactor(#2401): source task blocks from the canonical sectionizer The module carried its own copy of the <task>-block grammar - a fourth hand-rolled mirror of the one markdown-sectionizer owns. verify.cts keeps its copy only because it needs the type= attribute the canonical helper discards; this module never reads that attribute, so it can share the owner outright instead of adding a test around a copy. extractAutomatedCommands now takes task bodies from extractTaggedBlocks and the out-of-task remainder from stripTaggedBlocks. A task-grammar parity test pins the attributed task-name set against the canonical helper across six awkward task shapes. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#2401): extract agent-file overflow to references and repair the property arbitrary The remote matrix run came back red with 19 failures, four root causes: - agents/gsd-plan-checker.md and agents/gsd-planner.md both blew the 49152 agent cap. Their bodies move to gsd-core/references/, leaving @-reference stubs, per the documented overflow pattern. - The new checker dimension invoked gsd_run before the canonical preamble that defines it. The call is deleted outright: plan-phase.md already runs the probe and hands the result in as {VERIFY_PATHS}, so the dimension consumes that rather than re-running anything. - fc.fullUnicodeString does not exist in fast-check 4.8.0. Replaced with fc.string({ unit: 'binary' }), which covers the same 0000-10FFFF range. - Three runtime-loaded files grew; acknowledged in the existing ack fragments that already own those bare filenames, since two ack sources may never name the same path. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test(#2401): regenerate golden install-tree fixtures for the new references Adding two files under gsd-core/references/ changes what the installer emits into every runtime's tree, so all 19 golden install-parity fixtures went stale. Regenerated with npm run gen:install-tree; the delta is exactly the two new reference paths per runtime, no removals. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#2401): backfill changeset pr number to 3678 * fix(#2401): treat ~ as a home expansion only at the start of a path Windows CI caught this on both shards; the Linux-only remote matrix cannot see it. The dynamic-path refusal rejected ~ anywhere, and a GitHub Windows runner's tmpdir is an 8.3 short name - C:\Users\RUNNER~1\AppData\Local\Temp - so a valid absolute Windows path came back unresolvable/dynamic_path. This was a production bug, not a test artifact: any Windows user whose project path carries an 8.3 short name, or any literal ~, silently lost the probe entirely - every command degrading to unresolvable with no explanation. ~ is a home expansion only at the start of a path; elsewhere it is an ordinary literal. The check is now split: $, backtick, *, ? and newline stay refused anywhere (substitution and globs, and the glob characters are illegal in Windows path components regardless), while ~ is refused only leading, tolerating one leading quote since the check runs before quote stripping. The prior tests only caught this on Windows because only Windows puts a ~ in tmpdir. Four new tests pin it on every platform via a fixture directory literally named RUNNER~1. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-19 15:21:15 -04:00
Tom Boucher	1bf73d957b	enhance(#2295): record the resolved model per reviewer in REVIEWS.md frontmatter (#3649) ... * test(#2295): failing-first coverage for per-lane resolved-model recording * feat(#2295): record the resolved model per reviewer lane * docs(#2295): document the recorded reviewer model and its provenance * fix(#2295): refuse control characters in a recorded model value * test(#2295): correct watermark assertions for the widened mark shape * fix(#2295): anchor the role-manipulation injection pattern at a word boundary * feat(#2295): record the applied reasoning effort in the model value * chore(#2295): backfill changeset pr number * chore(#2295): restore em-dash in changeset body --------- Co-authored-by: sim <sim@local>	2026-08-19 08:59:37 -04:00
Tom Boucher	1adf6d2245	fix(#3620): point the docs at files that actually exist (#3658) ... * fix(3620): point the docs at files that actually exist docproof found 34 stale references; the reporter hand-read all 34 and reported the 8 that are real, explaining why the other 26 are deliberate (files the documents themselves label legacy or "superseded by", and one pre-Diataxis link label whose target still resolves). Those 26 are left alone — re-touching them would contradict the issue's own analysis. Every claim was re-verified against git ls-files at HEAD before editing. docs/INVENTORY.md said its roster is anchored by six drift-control tests. Five are gone (commands-doc-parity, agents-doc-parity, cli-modules-doc-parity, hooks-doc-parity in 5d8a8c4d; command-count-sync in fbf30792), so the sentence now names the one that exists. Whether one test is sufficient coverage is a maintainer question the issue explicitly declined to answer, so no new drift tests are proposed here. The four translations were a revision further behind, each naming a seventh test deleted in ae8bb707 that the English file had already dropped. All four now match. Renamed targets corrected in CONTEXT.md, VERSIONING.md, docs/CONFIGURATION.md and the update workflow. The new test names carry no issue-NNN- prefix, which is what lint-regression-test-names requires, so they are the correct targets. docs/TESTING-SUITES.md is the one that could cost somebody time: it INSTRUCTED contributors to add an acknowledgment to the legacy drift-ack file, which CONTRIBUTING.md says to never use. Rewritten from the real workflow — per-PR fragments under the drift-acks directory, and a spent base-side ack is re-armed by rewording that fragment's reason in place, never by adding a duplicate, since two sources naming one path is a hard error. docs/skills/discovery-contract.md's heading named a query module deleted in 11918dcc. The section was REMOVED rather than retargeted: its documented behavior (skip the deprecated root) is not what the surviving code does — skill-manifest includes that root marked deprecated:true — so retargeting would have documented something false. Found and fixed inline, same class: VERSIONING.md described an SDK bundling step the release workflow does not have (zero such mentions in that file); CONFIGURATION.md and four translations named a dead model-catalog triple collapsed by ADR-457. Dead config removed: the changeset lint's user-facing prefix list still carried two retired sdk entries. git ls-files -- 'sdk/*' returns nothing. No test pins that array. Left deliberately: the comment explaining the retired catalog path, the install regression test that reconstructs the old broken layout to prove it fails, and the generated test-timings cache. Each is a legitimate mention of a dead path, not drift. Note lint-removed-but-needed cannot catch this class: it diffs baseRef...HEAD, so it only sees files deleted in the change under review. These were orphaned by PRs that predate the lint. A repo-wide existence audit would need a suppression mechanism for the 26 deliberate mentions above; that is a feature, not part of this fix. Fixes #3620 * chore(3620): backfill changeset PR number (#3658) --------- Co-authored-by: sim <sim@local>	2026-08-19 01:54:01 -04:00
Tom Boucher	fe64704ace	enhance(#3588): add an opt-in commit_docs pre-commit hook (#3609) ... * feat(#3588): add an opt-in commit_docs pre-commit hook Final phase of epic #2292, scope narrowed to opt-in by maintainer decision: default-on installation and the bin/install.js wiring it would have required are explicitly out of scope. Enabling is an explicit verb call. The hook is written to the repo's real hooks dir resolved via git rev-parse --git-path hooks, so a linked worktree or submodule whose .git is a FILE works rather than getting a literal .git/hooks path. It refuses rather than overwrite a foreign pre-commit, refuses to delete one it did not write, and refuses outright when core.hooksPath is already set -- a written-but-ignored hook is worse than a refusal. Ownership is detected by marker presence, not byte-equality, so a user who appends a line does not make it unrecognizable. Deliberately NOT included: teaching cmdCheckCommit the per-phase commit_docs tier. #3587 was still unmerged when this landed, and implementing precedence against helpers that did not yet exist would have meant a second copy of the resolution chain -- the divergence class this epic has spent three phases fighting. That follows as its own change now that #3587 is on next. The ordering constraint is recorded in the design doc: this must not merge before #3587, or the hook would block a commit cmdCommit itself allows. * fix(#3588): teach the commit_docs guard the per-phase tier and -z paths Part 1, deferred until #3587 merged. cmdCheckCommit read only project-level commit_docs, so once #3587 landed, a phase with phase_commit_docs true under project false was ALLOWED by query commit and BLOCKED by this guard -- and the hook shipped in this same branch shells out to it. It now derives the staged phase via the single-owner detectPhaseNumberFromFiles and resolves through #3587's own resolveCommitDocsPolicy rather than a second precedence copy. Also fixes a proven false negative in the harm direction. git diff --cached --name-only C-style-quotes any path with non-ASCII or special characters, so a staged .planning/cafe.md was emitted as a quoted string, failed startsWith('.planning/'), and slipped past the guard entirely under commit_docs:false. Reading with -z and splitting on NUL removes the quoting at the source. The f.startsWith('.planning\\') branch was dead code under that read -- git emits /-separated paths on every platform -- and is removed rather than left implying coverage it never provided. The earlier C7 test pinned the buggy behavior as intended; it now asserts the file is detected and the commit refused. Self-caught: the commit-docs-guard verb was wired into the routers by this branch's earlier pass but missing from the top-level help listing. * test(#3588): replace try/finally with t.after, add negative-routing cases Standards review findings. CONTRIBUTING bans try/finally inside a test body outright -- it masks failures -- and B8 used one for worktree cleanup. Now t.after(), assertions unchanged. The new commit-docs-guard command family had zero negative-routing coverage, which CONTRIBUTING requires for any change to command dispatch. B11-B15 cover no subcommand, unknown, empty string, whitespace-only and a flag-shaped value, each asserting non-zero exit, a structured error, no stack trace, and -- the one that matters for a command that writes into a user's repo -- that NO hook is written in any of them. Those tests were verified to fail when routeCommitDocsGuard's else-branch is neutered, so they exercise the routing guard rather than any convenient error path. Also made two error() calls' control flow explicit with a return; they were safe only because error() is typed never two files away. * chore(#3588): backfill changeset pr number to 3609 * test(#3588): skip Windows-unrepresentable fixtures on win32 CI's Windows shards caught two of my own tests: fixtures whose filenames contain a quote and a backslash. Both are illegal on Windows -- backslash is the path separator, quote is invalid on NTFS -- so fixture creation failed before any assertion ran. Test-portability defect, not a production one. Those inputs cannot exist on that platform, so the guard has nothing to detect there. Both now check process.platform FIRST, before any fs or git call, and use t.skip() rather than a bare return -- a bare return registers as a PASS and would hide the gap it is meant to record. Each carries a comment saying the input is unrepresentable rather than unverified, so nobody later re-enables it. No padding added: the cafe.md case already exercises git's C-quoting path on every platform, since non-ASCII names are legal on NTFS. This is exactly the coverage the Linux-only remote matrix cannot provide, which the PR body already stated -- CI's Windows shards are what caught it. --------- Co-authored-by: sim <sim@local>	2026-08-18 00:25:32 -04:00
Tom Boucher	debeabd524	enhance(#3587): add a per-phase commit_docs override (#3601) ... * feat(#3587): add a per-phase commit_docs override Delivers epic #2292's second user story: commit an architecture phase's artifacts while execution phases stay local. commit_docs was project-wide and binary, so the only choices were all phases or none. Shape is a config dynamic key phase_commit_docs.<phase-id>, following the 14 existing dynamicKeyPatterns precedents rather than inventing a PLAN.md frontmatter spec -- which #2292 itself flags as becoming its own maintenance surface. Tier 1 resolves in cmdCommit, NOT in loadConfig: loadConfig has no phase context and is called by nearly every command, so threading one through it to serve a single caller would be a far larger blast radius for no gain. The phase comes from detectPhaseNumberFromFiles, which cmdCommit already computes for branch naming and which is already hardened against the #2539 project-code bug. Suppression by the per-phase tier returns its own reason rather than reusing skipped_commit_docs_false -- telling a user their project setting is false when it is true would be actively misleading. Additive; the two existing reason strings that agents/gsd-executor.md matches on are unchanged. The manifest's phase-id pattern is a hand-copy of PHASE_NUMBER_TOKEN_SOURCE because the manifest is hand-maintained JSON, so a behavioral parity test asserts both surfaces accept and reject the same token shapes. * fix(#3587): fold tests, close review findings, update reference docs Fold: the new tests were added as their own file, which required loosening a grandfathered lint-test-file-count bucket 5-to-6. A ratchet exists to go down only. commit-docs-bypass.test.cjs is the established commit_docs test home and already hosts two folded suites, so the tests fold there as a third block and the allowlist is reverted untouched. Standards review: CONTEXT.md and the test header both cited a phase-commit-docs-manifest-parity.test.cjs that never existed; a repo-wide sweep found a fourth stale cite in the schema manifest description. All four now name the real location. Spec review: the issue's Scope of changes named planning-config.md and git-planning-commit.md and neither was touched. Both now document the four-tier precedence and the new skip reason. Security review, minor and unproven: detectPhaseNumberFromFiles returns the FIRST matching path's phase, so a --files list spanning two phases resolves the override against whichever comes first. That helper is hardened and widely used, so it is not changed; the behavior is pinned by a named test and disclosed in the design and user docs. A pinned behavior is not a bug; an unpinned surprise is. * chore(#3587): backfill changeset pr number to 3601 --------- Co-authored-by: sim <sim@local>	2026-08-17 21:59:40 -04:00
Tom Boucher	3ab0007164	enh(#2875): materialization primitives — durable user-artifact staging and descriptor-authoritative agents (#3600) ... * fix(#2875): stage user artifacts durably across install wipes (#1874-F19) preserveUserArtifacts held user files only in an in-memory Map across the wipe, so any process death between preserve and restore lost them outright. Seven call sites, not the four the issue records. Three of them never called the helper at all - they open-coded the same read/wipe/write - so searching for callers under-counted by construction; the extra sites were found by sweeping for the pattern instead. The worst is the mainline install path, where the crash window spans the entire gsd-core tree copy rather than a single rmSync. Adds src/user-artifact-staging.cts: durable on-disk staging with a record written after the copies land as the commit point, plus recovery of orphaned batches on the next run - without recovery the staged bytes survive but the user's file is still gone, which would pass its own test while delivering nothing. Routes copyPreservingSymlink through installFs() so staging cannot bypass the install fs seam, and reunites its symlink-safety docblock with the function it documents. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(#2875): amend ADR-3574 with four claims disproved by implementation Implementing Phase 6 disproved four statements the ADR rests on. The central decision - no single materializer - is unaffected and stands. Corrected: decision 3 was already satisfied, so nothing was extracted; the agents-bypass runtime set omitted claude, kilo and opencode, and closing it needed three new pieces of descriptor contract rather than proceeding on its own terms; three of the four blockers the layout comment names were already stale; and F19 is seven call sites, not four. Records the generalizable lesson: the defect is the pattern of holding user data in memory across a wipe, not the helper, so searching for callers of the helper under-counts by construction. Also resolves the ADR's open question on USER_OWNED_ARTIFACTS membership, and notes that copyPreservingSymlink needed routing through the install fs seam before it could be reused. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#2875): close dangling-symlink blind spot and harden staging recovery An adversarial review found the F19 staging work shipped red and unsafe. Root cause, shared by two arbitrary-write findings: hasExistingSymlinkBetween missed dangling symlinks in both its root check and its per-segment walk, because it probed with existsSync, which is false for a link whose target does not exist. Fixing only the new module would have reused a guard that was itself blind. This guard protects the whole install tree. Recovery no longer throws: it degrades per entry and per file, so one bad batch cannot block the others. Previously an unrecoverable entry propagated out of the first statement of install and uninstall, before the cleanup that would have removed it - wedging the installer permanently. Partial fs adapters now throw on any omitted method instead of silently reaching the real filesystem, closing the trap that let a test poison list pass while real IO happened. Staged names must be flat, recovery refuses a dangling destination symlink, and a batch whose recovery genuinely failed is no longer swept - it was discarding the only durable copy of the file it had just failed to restore. Replaces three tests that could not fail, including the one labelled negative proof. Known limitation, documented not closed: concurrent installs sharing a staging key can still lose a batch. A real fix needs a cross-process lock. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * enh(#2875): make the descriptor authoritative for the agents kind Deletes the inline agent-staging loop in bin/install.js and the _DESCRIPTOR_AGENTS_RUNTIMES set, so every runtime materializes agents from its capability descriptor instead of an inline hostBehaviors dispatch. Closing it needed three pieces of contract the descriptor pipeline never had, all reducible to one missing input - per-agent resolution context: a frontmatter-extensions step for claude's effort and disallowedTools, per-agent model-override resolution for kilo and opencode, and a named branding converter for hermes, whose rewrite data was already declared. Seven runtimes were on the loop, not the six the design recorded - kimi-code was found by a golden fixture, not by analysis. claude-local and kimi-code both silently lost their agents mid-change; the fixtures caught both and the cause was fixed rather than the fixtures regenerated. A parity harness gates the migration: both pipelines over identical inputs, byte-identical output including filenames, per runtime. It is demonstrated red before being trusted. Surface and install paths converge for all seven, which also fixes surface previously writing no agents for these runtimes. Codex's config.toml strip stays put - it mutates host config, which no descriptor kind models. Also routes install-model-override-resolver and install-effort-resolver through the install fs seam. Both leaked real filesystem IO from the install call tree; the stricter adapter is what exposed them. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(#2875): record the agents-descriptor migration and correct the ADR count The _DESCRIPTOR_AGENTS_RUNTIMES allow-list no longer exists, so the host integration guide told readers to join a set that is gone. Replaces that with what is now true - declare an agents entry and it installs, on the surface path as well as install - and points anyone needing a per-agent transform at the three extension points rather than at a new inline branch. Corrects the ADR amendment: seven runtimes were on the inline loop, not six. kimi-code was found by a golden fixture going red, not by reading. That is the third short count this phase, all from enumerating by symbol or set membership when the thing that matters is a behavior. Adds the Changed changeset for the surface-path convergence. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(#2875): amend ADR-2866 - claude global always wrote agents on disk The claude row's global=[skills] described what capability.json declared, not what the installer wrote. bin/install.js's inline agent-staging loop was never scope-gated and never consulted the descriptor, so a claude --global install has always written agents/gsd-*.md. Phase 6 closes the gap by deleting that loop and declaring agents on claude's descriptor at global scope. On-disk bytes are unchanged - the golden fixtures did not move, which is the evidence that the descriptor, not the installer, was incomplete. #2218 is unaffected: agents are not trigger-bearing, so the wider row does not introduce a new shadowing case. Records the warning that an incomplete descriptor is invisible while a second code path silently does its work, and only surfaces when the two are forced into agreement. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#2875): close review findings across staging, agents and the parity harness Two independent reviews of this branch found defects the local gates missed. Security: a dangling symlink at a migration destination allowed writing outside configDir - the same class this change claimed to close, missed at the terminal write of the flow being added. The staging-root resolver threw as the first statement of install and uninstall, so a hostile symlink bricked both, and symlinked-configDir users lost uninstall as well as install; it now degrades instead of aborting. Recovery gained a source-side symlink check and now refuses a relative destDir, which resolved against cwd. Converter dispatch gained a runtime allowlist - lint-time validation stopped mattering once this branch promoted that dispatch from the surface path to real installs. Correctness: claude --local --minimal exited 1 because the minimal profile legitimately yields zero agents and the new path treated that as a failure. cline --local silently lost its agents - its descriptor declared none while the deleted loop wrote them unconditionally. The agents prune was widened to any gsd-* entry and destroyed user files it never owned. The parity harness, on which the migration's safety argument rested, drove a synthetic registry and never byte-compared the shipped descriptors; two of its trap rows could not fail. It now drives the real registry across 13 runtime-scope rows including kimi-code and cline-local, and its red-proof is demonstrated by corrupting a live capability.json. Three goldens that had encoded the cline regression as expected behavior were corrected. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#2875): close findings from both mandated review engines /security-review found the staging source-side walk honouring GSD_ALLOW_SYMLINKED_DEST, an opt-in documented as relaxing only the write destination. A symlinked files/ component dereferenced because copyPreservingSymlink lstats the leaf only, so an intermediate link is followed. The source walk no longer honours the opt-in; the destination check still does. /code-review spec axis found this branch had reintroduced its own bug: migrateLegacyDevPreferencesToSkill's new symlink refusal threw unguarded after the legacy dir was wiped and before the staged batch was restored, so a planted symlink bricked uninstall permanently and orphaned the batch. Refusal kept, abort removed. kimi-code local silently lost its agents, the same class as the cline bug, and the parity harness recorded that exclusion as intentional - the third test in this branch to pin a regression as correct. --minimal now creates an empty agents/ dir that never existed. Behaviour restored rather than softening the changeset, so its byte-identical claim stays true. Standards axis: try/finally removed from twelve test bodies, fast-check properties added for parseOwnerPid, boundary coverage at the grace window and the ancestor-probe depth, a parity assertion for the staging-root helper duplicated across two files, and the 8-deep config walk deduplicated. Records 60-review.json with every finding and disposition from five passes, including the smells left unfixed and why. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#2875): prune stale agents unconditionally in minimal mode The previous round stopped an empty agents/ directory being created when the resolved profile yields no agents. That was implemented by skipping the agents kind entirely, which also skipped its stale-agent prune - so a full to minimal downgrade left stale gsd-* agents behind. The deleted inline loop pruned unconditionally and only skipped writing. Those are three separate conditions, not one: prune always, write only when there is something to write, create the directory only when writing. Both call sites now run _removeGsdEntries before the empty-staged early exit. The symlink-escape guard moved with it, since the prune also touches dest. Codex .toml agents and the config.toml stanzas are cleaned again, and user-owned agents are still preserved. The agents/ directory is left in place after a prune empties it, matching every sibling kind - none of them remove the destination directory itself. Golden fixtures confirmed byte-identical: the prune is a no-op on a fresh install, so fixture generation is unaffected. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(#2875): document interrupted-install recovery for user-owned files The durable-staging fix is invisible to the user it protects. Someone whose install died mid-flight has no way to know USER-PROFILE.md was staged before the delete, that the next run restores it, or that recovery happens at the start of that run rather than in the background. Written as the task the user has - finish the interrupted command - rather than as a description of the mechanism, and states what it will not do: overwrite a file already present, or touch staging belonging to another install still running. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#2875): backfill changeset pr number Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test(#2875): assert the J8 model override without building a regex CodeQL flagged incomplete string escaping: the assertion interpolated the override value into a RegExp while escaping only forward slashes, which is meaningless in a constructor, leaving real metacharacters unescaped. The failure direction was the dangerous one - a metacharacter would have made the match more permissive, so the row would pass when it should fail. That matters here because J8 exists precisely because an earlier revision was a tautology; the rewrite reintroduced a different way for the same assertion to stop discriminating. Replaced with a line-wise exact match, so no regex is constructed at all. Swept the other test files this branch adds; no sibling instances. lint:ci passed on the original - lint-no-adhoc-regex-escape matches a full metachar-escape copy, so a single slash replace slipped under it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-17 17:25:53 -04:00
Tom Boucher	5f64d999dc	fix(#3586): warn when .planning/ is gitignored but still tracked (#3598) ... * feat(#3586): warn when .planning/ is gitignored but still tracked git ignore rules have no effect on files git already tracks, so a project that committed .planning/ before ignoring it keeps staging those files -- while commit_docs correctly resolves to false, which is exactly what makes the contradiction invisible. The probe lives in the SNAPSHOT BUILDER, not the rule: Rule.check may perform no ambient I/O (ADR-3180 8.1 rule 1, enforced by lint-planning-snapshot-bypass). buildPlanningTrackedField follows buildWorktreeHealthField's precedent -- injected execGit, bounded, degrading to UNREADABLE with a typed reason rather than throwing. W024 went inline instead only because no snapshot field carried its fact; that precondition does not apply here. W029 fires only on COMPLETE scope with ignored and tracked both true, so a degraded probe yields neither a finding nor a false all-clear, and the default project (tracked, not ignored) stays silent. The remedy is ADVISE-only -- --repair never untracks anything. * docs(#3586): document W029 and correct the health rule count CONFIGURATION.md documented the gitignore auto-detect without the caveat that ignore rules do not affect already-tracked files -- the very gap W029 exists to surface. Adds the caveat, the warning, its remedy, and why --repair will not act on it. CONTEXT.md's rule count was stale at 31 before this change (actual 32 through W028); corrected to 33 and pointed at the two other places the count is locked, so the next editor updates all three together. * fix(#3586): treat ls-files overflow as tracked, add CLI-level W029 tests Review findings. Security (minor, confirmed): execGit sets no maxBuffer, so Node's 1MB default applies to git ls-files. A .planning/ tree large enough to overflow it failed into git_list_failed and silenced W029 -- a false negative in exactly the large-history case most likely to have the real bug. Overflow is now treated as PROOF of tracking (the output was non-empty by definition) and resolves to tracked:true, scope COMPLETE, reason ok_truncated. Spec (major): test-matrix rows C1 and C2 were never implemented -- there was no CLI-level integration test at all, only rule-level ones. Both now drive the real validate-health dispatch and confirm W029 is reachable end-to-end. Known limit documented, not papered over: a deliberate git add -f under an otherwise-ignored .planning/ raises the same signal as the accidental case. There is no reliable way to tell them apart, the finding is advisory-only, and a heuristic that cannot actually distinguish them would be worse than the honest caveat. * test(#3586): update frozen health-doc counts and acknowledge health.md growth The remote matrix caught three gates that lint:ci does not cover. gen-health-docs.test.cjs froze a 35-row / 32-rule assertion; W029 makes it 36/33. Updated both the assertion and the test NAME, which embeds the counts -- a stale name is a lie even when the assertion passes. The second reported failure was the same assertion surfacing at describe-rollup granularity, not a distinct bug. emitted-attribution's growth arm needed an ack for the generated health.md. health.md was already named in 3309-health-docs-generated.json, and two ack sources naming one path is a hard error -- so a new fragment was not an option. That fragment's own history shows the pattern: #3309 created it, #2873 amended it in place for W028. Amended again for W029, with a note recording why this one file is amended rather than joined by a sibling. * docs(#3586): add the private-planning how-to and fix a wrong link docs/CONFIGURATION.md pointed 'Configure private planning' at how-to/configure-model-profiles.md -- an unrelated page -- and no private-planning how-to existed at all. Found while editing that section. The how-to test genuinely fires here: going private is four steps and crosses planning.search_gitignored, a setting owned by another concern, so a reference table structurally cannot carry it. The new page walks the whole sequence and leads with the step people miss -- .gitignore does not untrack what git already tracks -- which is the exact state W029 now detects. Also corrects 'artefacts' to 'artifacts' (repo house style is American). * chore(#3586): backfill changeset pr number to 3598 --------- Co-authored-by: sim <sim@local>	2026-08-17 15:56:46 -04:00
Tom Boucher	98ecb2ba8c	enhance(#2142): archive quick tasks at milestone close-out (#3592) ... * test(#2142): failing-first coverage for quick-task archival at milestone close-out * enhance(#2142): archive quick tasks at milestone close-out * fix(#2142): resolve review findings — readme injection, move/reset ordering, owned state write * fix(#2142): fold archival under milestone namespace, expose index IR, dedupe reset decision * test(#2142): assert archive-dir-relative summary path in index IR * docs(#2142): backfill changeset pr number to 3592 * test(#2142): skip newline-fixture injection test on windows (control chars illegal in path names) --------- Co-authored-by: sim <sim@local>	2026-08-17 14:51:00 -04:00
Tom Boucher	3c61b4a838	enh(#3565): sentinel/contract registry + check:contract-drift lint (#3571) ... * enh(#3565): sentinel/contract registry + check:contract-drift lint * fix(#3565): report artifact-row markers once and dedupe per marker * docs(#3565): backfill changeset pr number --------- Co-authored-by: sim <sim@local>	2026-08-16 12:59:08 -04:00
Tom Boucher	a4a02a7a01	enhance(#2874): return the executed plan and route install IO through a seam (#3568) ... * test(#2874): add failing-first gate for the executed-plan return Four rows from the matrix's red-first order. E3 pins the one early return, for the opencode family, where a void-shaped hole would otherwise survive unnoticed. E13 sweeps every runtime in the registry - enumerated from the registry rather than hardcoded, so a runtime added later cannot slip past. F2 proves absence of real filesystem contact rather than merely that the happy path ran, which is the difference between a complete seam and a partial one. G1 and G3 are the additive guard and must be green before and after. G3 deliberately leaves the two existing adapter test doubles untouched: if this change required editing them it would not be additive, and the acceptance criterion would be unmet. No production code. All 19 runtimes install without throwing today, so E3 and E13 fail on the undefined comparison alone. Refs #2874 * feat(#2874): return the executed plan and route install IO through a seam installRuntimeArtifacts returned void, so its correctness was observable only by re-reading disk. It now returns what it executed - per kind, per scope - including on the combinedFamilyInstall path, which was the one early return where a void-shaped hole would have survived unnoticed. Failure still throws rather than becoming an ok:false return, so control flow is unchanged for both existing callers. A best-effort cleanup that fails is still swallowed, but is now visible in the returned value rather than silently absent. The fs seam is ambient rather than threaded. Explicit deps through install-profiles and the 3000-line conversion module was impractical; the tradeoff, the synchronous-only re-entrancy assumption, the restore guarantee and the partial-adapter fallback trap are all documented at the seam. findInstallSourceRoot and its sibling stay unrouted by design - they locate the package's own source, not the install destination. readCmdNames keeps a second implementation because the standalone CLI that owns the original cannot require the compiled adapter without a build-order dependency on its own output. A parity test fails if the two ever disagree. Refs #2874 * chore(#2874): gitignore the new build artifact install-fs-adapter.cjs is tsc output from src/install-fs-adapter.cts, not a tracked source file. It was added to eslint's ignore list but not to .gitignore, so it landed as a tracked file - the third time this step of the new-.cts ripple has been missed on this epic. Refs #2874 * fix(#2874): close two seam leaks and correct a false comment A correctness review found the seam still leaked in two places, both subtler than the three already closed. readGsdCommandNames was routed when it should not have been: it reads the package's own commands directory, which a destination-fake is never seeded with, so under a fake adapter it returned an empty or wrong roster instead of failing loudly. It now reads real fs, matching the precedent already documented for findInstallSourceRoot. cleanupStagedSkills ran raw rmSync from a process exit handler, which is real filesystem work deferred past the point where withInstallFs has restored - the one thing the synchronous-only contract exists to exclude. Staging now captures the adapter that created each directory and cleanup replays it, so a real install cleans up exactly as before and a fake-staged path never reaches the real filesystem. Also corrected a comment claiming the migration reads were an unrouted, untested residual gap. They are routed and exercised; a comment understating the seam is as corrosive as one overstating it in a module whose trust rests on being honestly documented. Refs #2874 * test(#2874): migrate the exemplar group and cover the matrix AC3's exemplar migration lands in place: the qwen install group now asserts skills and agents destinations from the returned plan in one deepStrictEqual instead of probing the filesystem for each. Nine facts the old probes established were enumerated first. Two moved to the value assertion; seven were retained deliberately - per-file SKILL.md existence, the VERSION file written outside this function, the manifest content, and the post-uninstall absence checks all sit outside the plan's per-kind contract. A migration that quietly asserts less looks like a win and is a regression, so the enumeration is the guard rather than the line count. Also implements the rest of the matrix: the executed-plan shape, adapter failure modes, the security-boundary rows including a fake that cannot certify an install the real filesystem would refuse, cleanup visibility, and two seeded property tests. Only the two external CI gates are left unticked, because self-certifying them would be a claim rather than a check. Refs #2874 * fix(#2874): restore streaming hashes and derive F2 from the boundary rule The checkpoint found three things reasoning had missed. sha256File had been converted from raw-fd streaming to a single readFileSync on the assumption that GSD artifacts are never large. A test named for exactly that contract already existed and went red. Streaming is restored, now routed through the adapter, which gains openSync, readSync and closeSync. The contract was the specification; the assumption was not. Three existing tests inject faults by monkeypatching real fs. They broke because mkInstallTempDir stopped calling real mkdtempSync, not because of any binding subtlety - the real adapter was already late-bound. It now calls the real function when no fake is injected, so a monkeypatch applied after import is still seen and the additive contract holds. F2 poisoned real fs by method, so a deliberately unrouted package-source read failed a correct design. It now poisons by path: destination IO is forbidden, package-source IO is allowed and positively asserted. The claim was always zero real destination IO, and the test now derives from that rule instead of coincidentally matching it. Refs #2874 * docs(#2874): add the contributor how-to for plan-based test migration The phase gate caught a real gap. The docs plan was Reference plus Explanation only, and every CI check would have passed, because the docs-required lint only verifies that some file under docs/ moved. But this phase exists to demonstrate a pattern for follow-on work, and that work is other contributors migrating probing test groups. The sequence has two live traps - a partial fake silently falls back to real fs, and the seam is ambient and synchronous-only - plus one discipline nobody infers: enumerate the facts before converting, or you assert less and call it a win. The page carries the qwen migration's arithmetic, nine facts enumerated and only two converted, because a reader seeing only the diff would reasonably conclude the pattern is to replace probes wholesale. No locale mirrors: none of the four carries any contributor-only how-to, so a single translated file would manufacture parity rather than provide it. Refs #2874 * chore(#2874): backfill changeset pr number * test(#2874): normalize both sides of the G1 tree comparison G1 failed on Windows only, deterministically on both shards. The defect was in the test helper, not production. _computePathPrefix posix-normalizes the resolved config dir unconditionally, so on Windows the path embedded in every emitted SKILL.md body is forward-slash form. hashDirTree stripped against the raw backslash path from mkdtempSync, so the substring never matched and each install's unique temp suffix stayed baked into every file - all fifteen skill bodies hashed differently for two runs that had written identical bytes. Both sides are now normalized unconditionally rather than gated on path.sep, matching the rule this repo already records: backslash paths arrive on Linux too. Production code is untouched and was verified correct. Normalizing this away on the production side would have hidden a real portability bug if one had existed. Refs #2874 --------- Co-authored-by: sim <sim@local>	2026-08-16 02:48:24 -04:00
Tom Boucher	c5b83cb050	chore(#3560): delete two unreachable workflows, gate workflow reachability in lint (#3564) ... * chore(#3560): delete two unreachable workflows, gate reachability in lint discovery-phase.md and plan-milestone-gaps.md shipped to all 19 runtime install trees with no command, agent, or skill referencing them. plan-milestone-gaps' command was deleted by #2790 and the workflow was left behind; discovery-phase's own header claimed a caller in plan-phase.md's mandatory_discovery step, and that step does not exist — plan-phase.md contains zero occurrences of "discovery". docs/INVENTORY.md asserted discovery-phase.md was an alternate entry for /gsd-new-project. new-project.md never referenced it. The row and the matching note sentence are removed across all five locales rather than corrected. Adds rule 6 to lint-command-contract: every shipped workflow must be reachable from a loader, walking the transitive closure over the three reference shapes this repo uses. The closure seeds ONLY from commands/agents/skills, so a workflow that references only itself and a pair that reference only each other are both correctly reported rather than satisfying themselves; a visited set makes reference cycles terminate. The measure is a mention in a LOADER — docs/ and install-tree fixtures deliberately do not count, because scan.md proved a file can be documented and shipped while entirely unreached. Ships blocking, not report-only: #3561 is in this branch's base, so the tree reports 0 unreachable from the start. Closes #3560 * test(#3560): drive rule 6 end-to-end, sweep a stale allowlist, update ADR-0002 Review findings. Rule 6 had no end-to-end coverage: the tests exercised the pure closure with in-memory data, so the wiring — file collection, exit code, diagnostic — was unproven, and #3560's acceptance list explicitly wants a fixture showing the rule FAILS on a planted orphan. Adds an optional --root to lint-command-contract (default behavior unchanged) and four tests driving the real CLI through the process seam against a temp fixture: clean=0, planted orphan=1, orphan referenced only from docs/=1, orphan reachable transitively=0. The docs/ case is what pins the Goodhart defense — a mention outside a loader must not confer reachability. Deletes two tests that were byte-identical to a third and could not assert anything loader-specific, since the closure is source-agnostic by design; that distinction lives in the lint script's file collection and is now covered above. Removes a stale ALLOWLIST entry for discovery-phase.md in planner-language-regression — the exact sweep-miss class rule 6 exists to catch, found in the PR that adds the rule. ADR-0002 described five per-file frontmatter checks; rule 6 is a repo-level reachability graph, so the Decision section now says so. Refs #3560 * test(#3560): cut the bug-3298 test pin on the deleted plan-milestone-gaps workflow The remote runner went red with four failures: tests/phase.test.cjs asserted the plan-milestone-gaps workflow exists and checked its mkdir patterns, so deleting the file broke the test that pinned it. This is the fence the epic describes — the content-sync test IS what keeps an unreachable file alive — and cutting the coupling is what makes the deletion safe. Removes only that arm. The bug-3298 block guards three workflows against phase-dir prefix drift; the import and add-backlog arms and both shared mkdir-pattern helpers are untouched. Worth recording where the sweep failed: my reachability walk covered commands, agents, skills, gsd-core and docs, and lint-removed-but-needed covers .github/workflows, gsd-core, docs and package.json. Neither looks at tests/, so a test-pinned deletion is invisible to both and surfaces only on the remote runner. The how-to added by this PR names that gap explicitly so the next deletion searches tests/ by hand. Refs #3560 * docs(#3560): add a how-to for resolving unreachable-workflow findings * chore(#3560): backfill changeset pr number to 3564 --------- Co-authored-by: sim <sim@local>	2026-08-15 23:29:30 -04:00
Tom Boucher	1591454357	feat(#3409): reject shell guards that cannot observe their own failure arm (#3558) ... * test(#3409): failing-first regression tests for unreachable shell guard arms Drives the three live defects fail-first, executing the shipped workflow snippets rather than a re-typed copy: - G1/G2 plan-phase.md Walking Skeleton gate reads `--pick summaries_total`, a field that does not exist, so PRIOR_SUMMARIES is always "" and the gate has never fired (#3365). G2 is the load-bearing negative-space case: it rejects a fix that treats "no answer" as "zero" and fires unconditionally. - G3 plan-phase.md PHASE_REQ_IDS resolves "" instead of the TBD sentinel on a phase with zero requirements. - G4 complete-milestone.md's bare `cat <glob>` blocks on stdin under a nullglob left set by an earlier block (measured hang). Skipped on Windows for G4 only: the FIFO-blocked-stdin mechanism is POSIX only, and a weakened assertion there would pass vacuously. Refs #3409 * fix(#3409): make nine shell guards observe their own failure arm `--pick` coerces a missing field to empty string and exits 0, so the `|| echo <default>` fallback after it fires only on a verb typo, never on the field absence it was written for. Nine sites relied on that arm. - plan-phase.md walking-skeleton gate: `--pick summaries_total` names a field that does not exist under any flag combination, so the gate has never fired on any project (#3365). Repointed at the existing single owner, `phases.list --type summaries --pick count`, which returns a real integer in every case including a project with no `.planning` directory. No new counter is added: a second one would duplicate the ownership ADR-3180 Decision 1 forbids. The gate now fires only on a literal "0", so an unanswerable query fails safe instead of entering skeleton mode. - plan-phase.md phase_req_ids: now falls back to the documented TBD. - The remaining seven convert to an explicit empty test. - complete-milestone.md read all phase summaries through a bare `cat <glob>`; under a nullglob left set by an earlier block that is zero operands, so cat blocks on stdin. Guarded with the array shape the #3300 fix already established in review.md. Refs #3409 * fix(#3409): guard eleven more globs that defeat their own fallback arm The nullglob audit this issue asks for turned up the same class in files #3300 never touched. - Eight bare `cat <glob>` reads (transition, complete-milestone, planner x4, verifier, phase-researcher). With nullglob set that is zero operands, so cat reads stdin and blocks; measured rc=137 at 3s. - Three `ls <glob> || echo "<message>"` sites (session-report, review-backlog and its generated skill). nullglob makes ls succeed listing the cwd, so the message never prints and the user gets a directory listing instead. Guarded with `[ -e "${_ARR[0]}" ]` rather than `[ ${#_ARR[@]} -gt 0 ]`. The count form is correct only when nullglob is set, and six of these seven files never set it: without it the array holds the unmatched literal pattern, so the count is 1 and the guard passes wrongly. `-e` is correct in both worlds. review.md keeps its count guards — that block sets nullglob two lines above them. skills/gsd-review-backlog regenerated from commands/, never hand-edited. Refs #3409 * feat(#3409): add the unreachable-shell-guard drift lint A sibling of lint-planning-prompt-drift.cjs, consuming the shared scripts/lib/drift-scan.cjs rather than copying it, wired into lint:ci. Both detectors are one shape — a fallback arm defeated by a legitimate success-on-empty: - Detector A: `--pick` and `|| echo` on one line. `--pick` is the discriminator because "missing field renders empty at exit 0" is a documented CLI contract, not a heuristic. A rule keyed on gsd_run matched 111 lines, ~132 of them legitimate, and was rejected. - Detector B: `cat <glob>` in command position, and `ls <glob>` whose exit code feeds a real fallback or an if/while head. Informational `ls <glob>` whose stdout is consumed (97 sites) and `|| true` failure suppression (~15) are not guards and never fire. Shrink-only ratchet keyed on (file, trimmed text) with a per-pair count, POSIX-normalized unconditionally so Windows CI cannot report everything fresh and stale at once. Ships with a ZERO-entry baseline: every site it can find is fixed. Exemption is the per-line `# gsd-scan-ignore: #NNN` marker whose reason must name an issue or URL; a malformed reason reports a distinct error rather than silently exempting. No file allowlists. ADR-3409 records the invariant, the measurements behind both detectors, and why the upstream `--pick` contract fix belongs to #3473. Refs #3409 * fix(#3409): resolve review findings — typed surface, sanitized reports, tighter marker Standards axis (blocker): the guard's tests asserted on human-readable stdout/stderr and on free-form baseline-load prose, which CONTRIBUTING prohibits by name. Added the typed surface it prescribes instead of weakening the tests: a frozen REASON enum, a --json report mode, structured loadBaseline errors, and a test locking Object.keys(REASON) so a new reason stays three coordinated changes. Security axis: sanitizeForReport covered every violation field but not the baseline-load error path, which embeds raw JSON.stringify output -- that escapes nothing above 0x1f, so bidi and C1 controls reached CI logs unfiltered. Routed through the sanitizer at the output seam. Security axis: the scan-ignore marker accepted `#0` and a bare `http://`. Tightened to a positive issue number and a URL with a host. This diverges deliberately from the sibling in tests/commit-files-pathspec.test.cjs, whose looser form was copied verbatim; the header now records the divergence. Security axis: G4 built its FIFO with `mktemp -u`, reserving a name without creating it. Now created inside a `mktemp -d` directory. Spec axis: ADR-3409 claimed a ninth site landed after the issue was filed. git blame disproves it -- all nine predate it; the issue's hand count missed one. Corrected. The design and test matrix still specified B9 as a FLAG after implementation reversed it to PASS; both now record the reversal and why. Refs #3409 * docs(#3409): add the how-to for resolving unreachable-guard findings Reference and Explanation are carried by ADR-3409; this is the task-oriented quadrant CI cannot check for. The page exists mainly for one thing the lint structurally cannot catch: both `[ -e "${_ARR[0]}" ]` and `[ ${#_ARR[@]} -gt 0 ]` remove the glob from the command and therefore both pass, but the count form is correct only when nullglob is set — and nullglob is usually set in a different block of the same file. A reference table cannot carry that; a how-to can. Also documents the reason codes, so a reader can tell "nothing to report" from "could not look". No tutorial: this is a gate inside an existing CI loop, not a new entry point a newcomer starts from. Refs #3409 * fix(#3409): bring the touched prompt files back under their size gates The remote run was red on 14 tests, all size/attribution, none of them the regression suite. - agents/gsd-planner.md was 194 chars over a 49152 cap enforced by four separate tests, each of which says the remedy is extraction, not a bump. It had 41 chars of headroom before this branch. Its `## Checkpoint Types` section was an unlinked, condensed duplicate of references/checkpoints.md, which already carries all three types and their XML shapes; the section now points there and keeps the three names and percentages inline. Net -969, margin 1010. - gsd-core/workflows/execute-phase.md sat 2 chars under a comfortable margin assertion. Dropped the AUTO_MODE default: the `|| echo "false"` it replaced was unreachable, so the value was already sometimes empty on next, and its only consumer compares against `true`. Net -16. Left plan-phase.md's AUTO_CHAIN default alone -- that file names an explicit `false` branch, so empty would match neither branch. - Acknowledged the seven prompt files that genuinely grew, one specific reason each. Five of those paths were already claimed by spent fragments identical to next, which blocks a second source naming the same path; removed just the colliding key from each, deleting the two that this emptied. Refs #3409 * test(#3409): extract the whole PHASE_REQ_IDS block, not just its first line G3 failed on the remote runner with '' !== 'TBD'. The test was wrong, not the workflow. The shipped contract is now two consecutive lines -- the capture and the `${PHASE_REQ_IDS:-TBD}` default -- but the helper's `^PREFIX=.*$` regex returns only the first match, so the test executed half the contract and correctly observed the empty string. Renamed to extractAssignmentBlockFor and taught it to consume the contiguous run of lines sharing the prefix. The assertion is untouched: TBD is the right expectation, and weakening it to accept the empty string would have reinstated exactly the class this suite exists to catch -- a check that cannot observe the thing it is checking. extractFencedBashAfterAnchor is unaffected: it is fence-delimited rather than line-anchored, so G1/G2/G4 still capture their full blocks. Refs #3409 * chore(#3409): drop a spent ack fragment that collided on complete-milestone.md #3458 landed on next while this branch was in flight and its fragment claims complete-milestone.md, which this branch also grows. Two ack sources may never name the same path. Its entry is spent: the +9163 it explains is already absorbed at base, so it can no longer clear anything, and the checker's own guidance for spent entries is to delete them. Removing the key emptied the fragment, so the file goes too -- an empty one signals nothing. Refs #3409 * chore(#3409): backfill changeset pr number 3558 * test(#3409): hoist a regex subject out of exec() to clear the injection scan CI's prompt-injection scan flagged `MARKER_RE.exec('# gsd-scan-ignore: ...')`. The pattern `exec[[:space:]]*\(["']` is receiver-blind on purpose, so it catches `require('child_process').exec('...')` -- and the scanner's own header records that RegExp.prototype.exec is collateral, to be handled by its allowlist. Allowlisting the file would blind it to the real exec vector permanently, so the subject is hoisted into a const instead: same assertion, scanner left at full strength, no security surface widened. Refs #3409 --------- Co-authored-by: sim <sim@local>	2026-08-15 21:09:05 -04:00
sim	147856040b	fix(#2873): close review findings across fences, sanitizer and docs ... Isolated security review found resolveSpecRootReference's fence tracker toggled on any delimiter, so a backtick fence could be closed by a tilde one and an include in the gap was rewritten inside a code block. Fixed by reusing scanFencedBlocks - the canonical engine already behind stripFencedCode and extractFencedBlock - rather than carrying a fourth copy of fence detection, which also closes the duplication the standards review flagged. sanitizeForRender now strips combining marks and zero-width characters alongside the ANSI, control and bidi classes it already handled. Adds the C, E and F matrix rows the spec review found missing, including installer-level coverage that spawns the real install rather than calling the report builder. Ships the how-to, the reference and command docs in five locales, the changeset, the inventory and glossary entries, and regenerates health.md for the new W028 rule. Refs #2873	2026-08-14 23:48:39 -04:00
Tom Boucher	6badb839a0	fix(#3514): deny internal fetch hosts; disclose unverified integrity (#3516) ... * test(#3514): add failing-first denylist and integrity suites * fix(#3514): deny internal fetch hosts; disclose unverified integrity * docs(#3514): trust-model, glossary, and changeset entries * fix(#3514): scope v6 checks to literals; exact pin kinds in prompt * chore(#3514): backfill changeset pr number --------- Co-authored-by: sim <sim@local>	2026-08-14 21:34:29 -04:00
Tom Boucher	fba7c90327	chore(#3484): adr-0174 behavior carry-forward amendment and merge gate (#3507) ... * chore(#3484): adr-0174 behavior carry-forward amendment and merge gate * chore(#3484): regen example context index for new ruleset predicates * chore(#3484): review fixes - amendment heading per contributor-standards, helper-based fixtures --------- Co-authored-by: sim <sim@local>	2026-08-14 19:31:48 -04:00
Tom Boucher	6dbc124018	enhance(#3180): the sibling validators share one envelope and one owner — Phase 12 (#3407)	2026-08-13 11:31:16 -04:00
Tom Boucher	96a82bbffb	enhance(#3245): report the detected host runtime in init (#3307) ... * test(#3245): failing-first coverage for host runtime detection in init Locks the behavior epic #2313 Phase 5 must produce before any of it exists: init reports the detected host, explicit GSD_RUNTIME and config runtime still outrank detection, non-Codex sessions are untouched, and nothing is ever written to shared defaults (#2297). * enhance(#3245): report the detected host runtime in init init reported agent_runtime: claude inside a Codex session, and resolved agents_dir to the Claude agents root with agents_installed: true — a spuriously healthy triple. Runtime identity was only ever read from GSD_RUNTIME or an explicit runtime in .planning/config.json. Adds a detection rung beneath both explicit sources, in a new pure module. Codex is identified from its own documented session environment (CODEX_SANDBOX / CODEX_SANDBOX_NETWORK_DISABLED), else an explicitly exported CODEX_HOME whose config.toml exists. The default ~/.codex is never probed: that file exists on every machine that has run Codex, so probing it would misreport other runtimes' sessions. resolveRuntime keeps its exact contract and all 71 dependents, including formatGsdSlash command-style emission; only withProjectRoot consumes the new rung. Nothing is written on any path (#2297). Explicit config still wins (#2517). * fix(#3245): make the parity guard real and single-source the marker Four independent review passes found the generative-fix-divergence guard was vacuous: it asserted agreement at the one input where inferPreferredRuntime and detectHostRuntime do not differ, so it could not fail. It now pins the actual divergence point (CODEX_HOME set, config.toml absent) and records that the asymmetry is deliberate. The config.toml marker is now single-sourced from update-context.cts and imported, rather than carried independently by two surfaces. tests/helpers.cjs now scrubs CODEX_SANDBOX and CODEX_SANDBOX_NETWORK_DISABLED: GSD reads them, so an ambient Codex session would otherwise make the non-codex control test fail non-deterministically. Also: detection is throw-safe end to end rather than only around the fs probe; the Windows-join test is replaced with one that can actually fail (trailing-separator, catches hand-rolled concatenation); the #2297 no-write proof now wraps resolveReportedRuntime, the function that ships, across all three ladder outcomes. * chore(#3245): backfill changeset pr number --------- Co-authored-by: sim <sim@local>	2026-08-10 09:48:26 -04:00
Tom Boucher	d28ab7c8f7	enhance(#3243): sync installed codex .toml model/effort to the passive posture (#3296) ... * feat(#3243): sync installed codex .toml model/effort to the passive posture Implements ADR-2313 D7, and owns the Codex .toml typed IR that Phase 1's review assigned to this phase. The IR exists for a structural reason, not tidiness: this phase has to PARSE these files, and a parser kept bug-compatible with a separate renderer is the generative-fix-divergence shape this epic already dealt with once for the model predicate. So Phase 2's parsing MOVES here rather than being copied — agent-install-check now imports it, and its test file passing unchanged is the proof the extraction altered nothing. The load-bearing property is byte-identical round-trip: render(parse(x)) === x. Without it a sync silently reformats a user's file — line endings, key order, BOM, trailing newline — turning a two-line repair into a whole-file diff in their dotfile repo. The IR keeps original lines and removes targeted ones rather than reconstructing from parsed fields, which is what makes that property hold. It also reconciles a real contradiction between Phase 2 and ADR-2313. An unterminated developer_instructions block: the reader excludes the rest of the file, deliberately failing toward a false positive, because misreading prose as a pin only wastes a user's time. The writer must refuse, because proceeding on a malformed document rewrites it. A false positive is the safe direction for a reader and the dangerous one for a writer. So the parse reports the fact and the two consumers branch on it — one parse, one truth, two policies, instead of two parsers that agree today. The sync leaves a legal real-Codex pin and its coupled effort untouched, reported skipped rather than synced; strips a stale Anthropic or tier model and an orphaned effort; keeps dry-run as the default; refuses any file whose parse fails; and skips symlinks exactly as the Claude path already did. The Claude path itself is byte-identical. PARSE_REASON.NO_HEADER from the ADR's illustrative snippet is deliberately not implemented — a missing header is legal, not an error, so it would be a dead enum member that the enum-lock test then pins. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3243): preserve per-line endings and make the codex write atomic Two findings from an isolated review, both in the write path. BLOCKER: mixed line endings broke the byte-identical round-trip. `eol` was a single whole-file flag and split(/\r?\n/) discarded each line's own terminator, so render re-joined with ONE style and normalized every line — even with zero strips performed. A file with one CRLF line and the rest LF came back fully converted. That falsified the A14 guarantee, violated the design's "must not silently rewrite every line", and made the CONTEXT.md glossary claim wrong. It was untested because A12 and B15 only cover PURE CRLF; no mixed-ending fixture existed anywhere. Fixed by keeping each line's terminator alongside its content, so render is a plain concatenation and a strip removes only the target line and its own terminator. `eol` survives as informational metadata that render never reads. Seven fixtures added for the paths nothing exercised: mixed endings unmodified and with a strip, a lone \r, a file ending on the block's closing ''' with no newline, multiple trailing newlines, a BOM-only file, and an empty file. MINOR, but it contradicted this phase's own contract: the write was in-place open-truncate, so a failure between truncate and completion leaves a truncated .toml — exactly what ADR-2313 says must never happen. The Codex path now writes a sibling temp file and renames over the target, which is atomic on one filesystem, with cleanup on failure. It uses the repo's existing retryRenameSync rather than a hand-rolled rename, and deliberately NOT platformWriteSync, whose normalizeContent would mangle the very CRLF and trailing-newline bytes the round-trip property exists to preserve. The Claude path keeps its in-place write untouched. It has the same shape, but changing it is not this phase's business and its tests must stay byte-identical. B20 previously mocked writeFileSync to throw BEFORE touching anything, so it proved nothing about a mid-write failure — its passing comment was true only because of how the mock was built. It now performs a real truncated write wherever writeFileSync is called, catching both the naive direct-to-target path and the new temp path, and asserts the target is byte-identical afterwards with no stray temp file left. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3243): preserve the trailing-newline state when stripping a last line Caught by B17, one of this phase's own tests — the suite working, not a test problem. Content is reconstructed as the concatenation of lines[k] + terminators[k], so a file with no trailing newline has '' as its last terminator. removeLine spliced out both arrays at the same index, which is right for a middle line but wrong for the last one: it dropped the empty terminator and left the PREVIOUS line's newline in place. A file ending `...\nmodel = "sonnet"` with no trailing newline came back as `...\n`, gaining a newline the user never wrote. The new last line now inherits the removed line's terminator, so a removal leaves the file exactly as if that line had never been written. Removing the only line yields an empty file rather than a stray terminator. Both stripModel and stripReasoningEffort funnel through the one removeLine, confirmed rather than assumed, so a single fix covers both — including the row-B7 shape where a stale model and its orphaned effort are removed in sequence and the second removal targets the last line. Two of the four new cases are honestly not red-first and say so in their comments: removing a last line that HAS a trailing newline only exposes the bug under mixed EOL, since uniform files coincidentally have equal terminators on both sides; and removing the only line already degenerated correctly through Array.slice. They are kept as guards for the new branch rather than dressed up as catches. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(#3243): document the codex repair path and close the loop How-to: the Codex-400 entry added in Phase 2 told users to re-run the installer, because that was the only repair available then. It now leads with `effort sync` and keeps the reinstall as the alternative, with the reason to prefer one — a reinstall regenerates the agent files wholesale, so anyone who hand-edited theirs loses those edits. Detect, preview, apply is now one continuous path in one place. Reference: docs/COMMANDS.md had no `effort sync` entry at all — the same gap `validate agents` had in Phase 2, found the same way. The entry documents BOTH runtimes, because the command genuinely forks on runtime and describing only the new half would misdescribe it. The write flag is `--apply`. The design doc and test matrix both said `--no-dry-run` throughout, which does not exist — verified against the actual arg parser in gsd-tools.cjs before writing. Documenting a flag that does not exist is worse than documenting nothing, because it fails at the moment someone needs it. Both surfaces state that only the targeted lines are removed and every other byte is preserved. That is a user-visible guarantee rather than an implementation note: it is the difference between a two-line diff and a reformatted file in someone's dotfile repo, it is what the IR's round-trip property exists to deliver, and writing it down makes it a contract a future change has to break knowingly. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3243): inherit the trailing-newline state, not the line ending style My previous rule was subtly wrong and this phase's own test caught it. "The new last line inherits the removed line's terminator" copies the removed line's STYLE as well as its presence. A26 uses mixed endings on purpose — line one terminated \r\n, the model line terminated \n — so inheriting silently rewrote line one's ending to \n. That is precisely the defect class the mixed-EOL blocker fix existed to eliminate, reintroduced one layer down by the fix for it. The correct rule inherits the EMPTINESS only. If the removed line had no terminator, the new last line loses its own, preserving "this file has no trailing newline". Otherwise the new last line keeps its own terminator: it is already a newline, and already the right style for that line. A26's assertion moved too, and that deserves saying plainly rather than burying: it previously encoded my wrong rule. Changing a test to match the implementation is usually the mistake, so it was checked from first principles instead — a file whose first line ends \r\n and whose last line ends \n, with that last line removed entirely, must be the first line with its own \r\n intact. The new expectation is what the user's file should actually look like; the old one was wrong. A29 adds the interaction nothing covered: the compounding case (strip a stale model, then its orphaned effort, the second removal landing on the last line) with non-uniform endings either side. The two fixes meet there and nothing exercised the meeting point. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3243): drop the phantom trailing line from the IR representation Root cause, not another patch on the removal rule. Three consecutive fixes there each surfaced the next issue, which was the signal that the data model was wrong. splitPreservingTerminators left a phantom empty final entry for any file ending in a newline: "a\nb\n" became lines ['a','b','']. So for the common case the real last content line was NOT the last array element, removeLine's isLastLine check never matched it, and every rule I gave was reasoning about the wrong element. What hid it: render was already a plain concatenation, so a phantom empty line with an empty terminator contributes nothing to the output. A14's byte-identical round-trip could never have caught it — the defect is byte-neutral until a removal shifts the index arithmetic under it. That is worth recording, because "the round-trip test is green" was exactly the reassurance that kept the search pointed elsewhere. The representation is now 1:1 — terminators[i] follows lines[i] and may be '' — with no phantom, verified across empty, no-trailing-newline, trailing-newline, blank-line and mixed-CRLF inputs. render stays a plain concat and needs no special cases. With the phantom gone the removal rule is correct as stated and finally applies to the genuinely last element. Consumers checked rather than assumed: the block-range detector and header scanner are agnostic to array shape, and Phase 2's reader uses its own independent split, so tests/agent-install-check.test.cjs is untouched and still passes unchanged. One test expectation was wrong and is corrected rather than quietly adjusted: A18 asserted a 7-element terminators array whose trailing '' was the phantom itself. It now asserts the six real terminators, which is what the invariant lines.length === terminators.length requires. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#3243): backfill changeset pr number (#3296) --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-10 01:31:15 -04:00
Tom Boucher	693f12ad56	refactor(#3187): give state field extraction one canonical owner (#3283) ... * refactor(#3187): give state field extraction one canonical owner stateFieldValue in state-document.cts becomes the single owner of the #1760 frontmatter-then-body fallback chain. The new whole-repo guard found 14 independent re-derivations where the epic scoped 5, all now routed through it: cmdStateSnapshot (11), cmdStatePrune (2) and smart-entry fmScalar (1). state validate was a gate that could not fail. Every warning it could emit sat behind a phase resolved without the frontmatter tier, so a STATE.md whose phase lives only in frontmatter skipped the drift scan entirely and returned valid:true. It also read unstripped content, letting a frontmatter status: key shadow the body field (#1255 class). Both fixed; output gains a scope field so could-not-look stops being output-identical to looked-and-clean. Verified on the remote runner. * docs(#3187): document the state validate scope field and its reason codes Adds docs/how-to/interpret-state-validate-results.md so a reader can tell nothing-to-report from could-not-look, updates the COMMANDS.md and USER-GUIDE.md entries, corrects the CONTEXT.md glossary overstatement about Current Position sole ownership, and drops the changeset fragment. * fix(#3187): close three drift-guard evasion shapes and test the refuse path The isolated adversarial review found the ladder detector was evadable by ordinary reformatting, not just deliberately: a member or computed operand (fm.key / fm[key]) missed the bare-identifier backreference, a swapped tier order missed a hardcoded number-then-boolean sequence, and a ladder wrapped across lines missed single-line detection. All three now caught, each with its own test plus a proven boundary control. The frontmatter-parse refuse path on the destructive complete-phase route was unreachable and therefore untested. It is now driven by an injected parse failure and asserts STATE.md is byte-identical after the refusal, rather than shipping untested defensive code on a path that rewrites user state. Verified on the remote runner. * fix(#3187): widen the drift guard to the prompt layer and disclose tier-2 changes The code-review spec axis found the guard's scan surface was src/ only, which is Decision 4(d)'s forbidden allowlist one directory wide - and it had a live miss: gsd-core/workflows/smart-entry.md tells an agent to read status from frontmatter or the body, a prose expression of this same chain. The surface now covers the prompt layer. That one site carries a permanent written exemption rather than a ratchet: it is the gsd-tools-is-down fallback, so it cannot call the owner by construction, and a ratchet would imply removable debt that does not exist. Two tier-2 output changes shipped undisclosed and are now named in the changeset and docs: complete-phase's idempotency guard consulting frontmatter, and the workstream inventory resolving frontmatter-only fields. docs/COMMANDS.md gains a state complete-phase entry, which it never had. Also records Amendment 5 on ADR-3180, extracts the duplicated frontmatter-parse block the epic's own thesis forbids, and re-points two assertions from free-form warning prose onto the structured drift object. Verified on the remote runner. * chore(#3187): backfill changeset PR number pr:0 placeholder replaced with the real PR number now that #3283 exists. --------- Co-authored-by: sim <sim@local>	2026-08-09 22:49:39 -04:00
Tom Boucher	4a1ed2531f	enhance(#3242): validate codex .toml model posture, not just presence (#3290) ... * test(#3242): failing-first suite for the codex posture health-check Specifies ADR-2313 D6 before the implementation exists, so the tests bind to the contract rather than to whatever the code happens to do. RED is established by construction, not by a remote run: checkCodexModelPosture and POSTURE_REASON are absent from the compiled lib today, so every row fails on the missing export. A remote checkpoint here would prove only that the function is missing, which is already known — so the run is deliberately deferred to the combined green checkpoint rather than spent proving a tautology. That makes the NEGATIVE PROOFS the rows that carry real signal. Every positive row passes even for a naive implementation that greps /model\s*=/ over the whole file. Six rows fail it: light-tier service_tier/model_verbosity decoupling (#774), hand-added keys, a commented pin, the model_verbosity key-prefix collision, the runtime no-op ordering, and the headline case — a literal `model = "sonnet"` inside the developer_instructions ''' block, which the emitter fills with agent prompts that discuss models constantly. Row 14's fixture was verified to discriminate before being written: a whole-file scan matches it and a header-slice scan does not. Without that check the test would pass trivially and prove nothing, which is the vacuous-test failure this epic has already hit repeatedly. Adversarial TOML fixtures are hand-authored against the real Codex shape rather than generated by generateCodexAgentToml, per #2371 — a fixture from the writer can only confirm what the writer already believed. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * feat(#3242): validate codex .toml model posture, not just presence Implements ADR-2313 D6. checkCodexModelPosture is a new sibling export, not a branch inside checkAgentsInstalled — that function carries 33 upstream dependents, cyclomatic 25, and sits in two traced process flows, so it is deliberately left untouched. It imports isAnthropicFlavoredModel from model-catalog, a genuine leaf. That is what Phase 1's constant move bought: agent-install-check is documented as pure read/verify and imports only leaves, so reaching the rule through model-resolver would have dragged config-loader into it. Reads liberally, judges strictly, and never guesses. Tolerates comments, key order, whitespace, CRLF, and a BOM; anchors on full key names so model_verbosity does not satisfy a `model` probe; treats extra hand-added keys as none of its business, since the check is a predicate on the two fields the posture owns rather than a whitelist over the document. An unreadable file becomes a named violation and the loop keeps going. The scan covers only the header slice — the lines before the developer_instructions ''' marker. The emitter writes agent prompts into that block and GSD's prompts discuss models constantly, so a whole-file scan reports violations for prose. This is the highest-risk defect in the phase and the reason its fixture was verified to discriminate before being written. The non-codex short-circuit runs before any filesystem call, so a stray .toml under another runtime is never inspected. Wired through cmdValidateAgents as an additive codex_posture key, so a violating install is visible from a command a user actually runs rather than only from a library nothing calls. Also fixes a test defect found while implementing: .gitattributes forces `* text=auto eol=lf` repo-wide, so the committed CRLF fixture was normalized to LF in the index — `git ls-files --eol` reported `i/lf w/crlf`, the working copy being stale pre-normalization bytes. The CRLF row was asserting against a file that could not survive a fresh clone. CRLF is now derived at runtime, which puts it under the test's control rather than git's, instead of adding a .gitattributes exception that fights a deliberate repo-wide policy and that anyone could re-normalize. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(#3242): document the codex posture check where a user will look Three quadrants, filed by where the reader actually arrives. How-to (recover-and-troubleshoot.md, under Install and update problems) is titled by the SYMPTOM — "If Codex agents fail to spawn with a 400 about an unsupported model" — and opens with the verbatim error string. Someone hitting this does not know the words "posture" or "ADR-2313"; they have a 400 in their terminal and will search for that. Reference (COMMANDS.md) had no `validate agents` entry at all, though sibling gsd-tools subcommands are documented. Adding user-visible output to an undocumented command and then linking to it from the new how-to would have left a dangling reference. The entry carries the violation-reason table, since the frozen POSTURE_REASON enum is the machine-readable contract a reader needs rather than the prose. Both surfaces state that presence and posture are separate verdicts — a missing agent lands in `missing`, never as a posture violation. That is a deliberate design decision and would otherwise be invisible to someone watching one command emit both. Explanation stays in ADR-2313, which already covers D6 and the liberal-parse/strict-judge boundary. Pointing at it beats duplicating it into COMMANDS.md and creating two copies to drift. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3242): close two false negatives in the posture scan Both found by an isolated reviewer and reproduced before fixing. Both made the check report clean when it was not — the worst direction for this function, since the how-to tells users an empty violations list means the install is posture-clean. A quoted TOML key was never matched. `"model" = "sonnet"` is legal TOML, but the key pattern required a bare identifier, so the pin was silently invisible. Bare, "double" and 'single' quoted forms now normalize to the same key name. The block marker was found by unanchored whole-content search and used to truncate the header. A `description` value merely containing the literal text `developer_instructions = '''` truncated the scan before a real pin, and a user who hand-reordered `model` to sit after the block — still legal TOML — was never scanned at all. Fixed by changing the strategy rather than the regex: find the block's range, anchored at line start, and scan every line OUTSIDE it. That covers both failures and is strictly more correct than truncation, while still never reading prompt prose. An unterminated block excludes the rest of the file, which fails toward a false positive — the safe direction, since misreading prose as a pin wastes a user's time while the alternative hides a real one. Also corrects two overclaims of mine. The how-to named "v1.11", a version that does not exist — package.json is 1.10.0 and unreleased — so it now describes the boundary by behavior and links the ADR. And the test matrix asserted that a naive whole-file scan "fails exactly rows 12,13,14,15,16,25"; the reviewer computed that rows 12, 13, 15 and 16 produce the correct result against that baseline too. They guard real but *different* mistakes, and the matrix now says which one each catches instead of attributing them all to the header-slice defect. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3242): skip symlinked agent files instead of following them Security review finding. The scan listed entries with readdirSync and read them with readFileSync, which follows symlinks — so a symlink in the agents directory pointing anywhere would have its contents read, and any line matching the model pattern echoed into cmdValidateAgents' output through the `value` field. A read-and-echo primitive on an arbitrary path. It needs write access to the agents directory, so it crosses no new trust boundary today. Fixed anyway, for two reasons. This repo already does it correctly next door: cmdEffortSync filters with lstatSync().isFile() and the comment "Skip symlinks — only write regular files to avoid clobbering symlink targets." Being inconsistent with a sibling in the same subsystem IS the defect. And Phase 3 (#3243) extends that same cmdEffortSync to WRITE these files. Establishing symlink-following as the house pattern for Codex .toml handling here would hand Phase 3 a worse starting point while it writes rather than reads. Skipped silently rather than reported, matching the sibling: a symlinked agent file is a structural install choice, which checkAgentsInstalled owns, not a model-content posture defect. An lstat that itself throws excludes the file rather than crashing the scan. That does narrow the guarantee slightly, so the how-to now says an empty list means every REGULAR .toml is clean, and tells anyone symlinking their configs to check the targets by hand. Claiming a clean bill of health over files the check declined to open would be the same kind of false confidence the two false negatives above produced. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#3242): backfill changeset pr number (#3290) --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-09 21:50:49 -04:00
Tom Boucher	b901d1e06f	feat(#1953): complexity-triggered refactor extension point (execute:post) (#3261) ... * test(#1953): failing-first suite for the complexity-triggered refactor hook 60 behavioral cases against src/complexity-trigger.cts, which does not exist yet: decision-point counting, the comment/literal stripping leak surface, threshold and jump-delta boundaries at limit-1/limit/limit+1, stable-anchor baseline semantics, and fs fault injection via mock.method. Two fast-check properties assert that stripping never manufactures a decision point and that comments and string literals are score-neutral. Also registers the refactor-trigger capability manifest (inert until refactor.trigger_enabled) and regenerates the capability registry and matrix. Verified RED on the remote runner before any implementation exists. * feat(#1953): complexity-triggered refactor extension point Adds the opt-in refactor-trigger capability. After a phase executes, an execute:post step measures per-function complexity for the files the phase touched and writes a scoped refactor proposal when a function crosses the configured threshold or drifts past its recorded anchor. Design notes worth carrying: - The signal is computed in-core (decision-point counting over comment- and literal-stripped source, Node builtins only) rather than via Memtrace or a shelled-out analyzer. The hook fires as a deterministic CLI, not an agent with MCP tools, and core takes no external dependencies — this is the only option a behavioral test can bind to. The metric sits behind a seam. - The baseline is a stable anchor, not a rolling value: set on first observation, moved only on disposition. A rolling baseline makes the delta the single-phase change, so a function creeping +2 per phase never trips a delta of 5 and the jump check adds nothing over the absolute threshold. - Strict mode records an open deviation window in the broken-windows ledger rather than declaring its own ship:pre gate. ship.md has no generic ship:pre gate dispatch — only two hardcoded branches — so a third gate of any kind would be declared and never evaluated. - The gate clears on the proposal being dispositioned, never on the score improving. A blocking complexity number is one an executor can satisfy by splitting a coherent function in two. execute-phase.md gains a generic execute:post step-dispatch contract; it previously matched only ref.skill == "code-review", so any other step registered there was declared and never run. The code-review branch is unchanged. Full rationale in ADR-1953. Closes #1953 * fix(#1953): close git option injection and symlink escape in the refactor hook Three findings from the isolated security review, all fixed inline. HIGH — changedFilesSince interpolated the --since value into a revision token placed before the -- separator. A -- only stops PATHSPEC parsing of arguments after it; git still option-parses what comes before. So --since '--output=/tmp/x' became --output=/tmp/x..HEAD, which git accepts as --output=<file> and uses to redirect diff output — an arbitrary write. Fixed with --end-of-options before the revision range plus a conservative ref validator. The validator deliberately permits ~ ^ @ { } because those are legitimate git REVISION syntax (HEAD~1, main@{yesterday}) as distinct from ref-NAME syntax; --end-of-options is the actual barrier. The doc comment asserting the trailing -- was sufficient was wrong and is corrected. MEDIUM — resolveConfinedPath confined by string prefix only, so a symlink committed inside the repo passed the check (its own path is under cwd) and readFileSync then followed it outside the root. Now lstat-checks for a regular file and skips anything else with REFACTOR_FILE_UNREADABLE, so one bad path skips one file and the run continues. LOW — the new execute:post dispatch contract showed the gsd_run example before the rule requiring ref.command be validated first. That prose is executed by an agent, so textual order is execution order. Reordered. Refs #1953 * fix(#1953): make the analyzer able to see TypeScript at all Found by running the shipped analyzer over its own source: it reported functions=1 for a 940-line module with 24 function forms. A return-type annotation or a generic parameter list made a function invisible — `function f(a): number {}` and `function f<T>(a: T): T {}` both detected as zero. Since gsd-core is written in .cts and the capability declares .ts/.cts/.mts analyzable, the feature silently found nothing in this repo's own primary language while reporting success. A safety net that reports "all clear" because it cannot see is worse than no safety net. All 98 tests passed over this, because every fixture was plain JS — the exact failure the test matrix's own "assert against the shape production uses" warning describes. Adds a TypeScript-shapes suite covering return types (including unions, generics, object literals and type predicates), generic parameter lists (constrained and defaulted), export/async/generator combinations, annotated arrows, class-method modifiers, and optional/ default/rest params — plus the two traps: an overload signature has no body and must not count, and `a < b && c > d` is a comparison, not a generic. Detection now reports 24/37/21 functions for the three source files, which matches a hand count exactly. Also from review: - The strict-mode ledger dedup identified entries by parsing a prose description string. That is banned by CONTRIBUTING's raw-text-matching rule and was a real bug: the "exactly one window per untriaged proposal" guarantee rested on prose matching, so rewording a description or editing WINDOWS.md by hand silently produced duplicates. Now matches structurally on kind + phase + file + line. - A property test asserted on the stripper's output text. Reframed to assert the same invariant through analyzeSource's score. - nextBaseline's `candidates` parameter has been dead since the anchor change; removed from the signature and all call sites. - Extracted the duplicated require-or-degrade and capability-check boilerplate. - ADR-1953's Implementation bullet still named a `refactor.ship-gate` in check-command-router.cts — a leftover from the design cut D6 rejects. That file is untouched and no such gate exists. Removed. Refs #1953 * fix(#1953): keep execute-phase.md under its byte ceiling; un-vacuum the large-file test Five of the seven remote-runner failures were one cause: the execute:post dispatch contract, written out inline, grew execute-phase.md 1876 bytes (93,400 -> 95,276) against a frozen PRE_PHASE6 ceiling of 93,600. A drift-ack does not clear that — tests/phase6-capstone-conformance.test.cjs and tests/fix-2285-claude-orchestration-wiring.test.cjs assert the file is literally under the cap. The contract now lives in gsd-core/references/loop-hook-dispatch.md, which already claimed to be the point-agnostic dispatch reference and already documented ref.skill and ref.agent. It gains the ref.command shape, its in-context validation rule, the advisory-by-construction statement, and a note that a point whose workflow hand-rolls one kind is not implementing this contract. execute-phase.md now defers to it in one line: 145 bytes of growth, 55 B of headroom under the cap. Better placement than the first cut — the reference was overstating its coverage, and this makes the claim true rather than duplicating prose next to it. Acknowledged by appending to tests/emitted-drift-acks/2930-*.json rather than a new 1953-*.json: two ack sources may never name the same path, and that fragment is already the accumulating ack for this file. Sixth and seventh failures: analyzesLargeFileWithinBounds tripped its own vacuity guard — the fixture generated ~480 KB against a `> 500000` assert, so the guard fired and the three assertions after it never ran. The test has been vacuous since it was written. The matrix row specifies ~1 MB, so N goes 8000 -> 20000 (1.17 MB, 17% margin) and the guard to > 1_000_000. Verified by reproducing the exact body against the compiled module: 1168888 bytes, 118 ms, all four assertions hold. Refs #1953 * fix(#1953): fold the execute:post step deferral into the existing resolve line The remaining two failures were one test: execute-phase.md carries a SECOND, tighter assertion than the 93,600 ceiling — `<=93400`, which is exactly its current size. The file cannot grow by a single byte. My previous fix got it under 93,600 but not under 93,400, so it still failed. ("H." in the report is just the parent describe of that same test, not a separate defect.) Rather than add a paragraph, the deferral now REPLACES the existing hook resolution line. It read: Resolve active step hooks from `EXECUTE_POST_HOOKS_JSON` where `kind == "step"` and `ref.skill == "code-review"`. which is the bug itself written down — only code-review was ever dispatched. It now reads: Dispatch each `kind == "step"` hook per @gsd-core/references/loop-hook-dispatch.md. For `code-review`: The following prose already begins "If no active code-review step hook exists", so it reads correctly and the code-review handling is untouched. Net effect on the file is -11 bytes: 93,400 -> 93,389, under the margin assertion rather than merely under the ceiling. That also removes the need for a drift-ack: the file shrank, so there is no growth to acknowledge, and the append to the shared 2930-*.json fragment is reverted. Leaving it would have shipped a claim of "145 bytes of growth" that is no longer true, on a file six other issues share. The test's own comment states the principle this ended up honoring: "the host loop must stay small — optional-feature detail belongs in the capability fragment, not the host workflow." Putting the dispatch contract in the reference rather than inline is that rule, applied. Refs #1953 * fix(#1953): keep the code-review hook literal the workflow test requires tests/code-review.test.cjs extracts the <step name="code_review_gate"> block and asserts it contains `ref.skill == "code-review"` verbatim. The previous commit replaced the line carrying that literal, so the token vanished and the test went red — a fair assertion: code-review IS the bespoke branch there and the workflow should still name it. Restored inside the same one-line deferral, which now reads: Dispatch `kind == "step"` hooks per @gsd-core/references/loop-hook-dispatch.md. `ref.skill == "code-review"`: 93,396 bytes — still under the `<=93400` margin assertion and 4 bytes below the base, so the file continues to shrink rather than grow. Because three consecutive runs were each reddened by a different assertion on this one file, this change was verified by sweeping ALL of them at once rather than one run at a time: every test under tests/ that reads execute-phase.md or references/loop-hook-dispatch.md was located by resolving its path constants, and each content/size assertion was evaluated directly against the working tree — 22 assertions, plus two real executions (gen-section-manifest --check, and emitted-attribution's full real-tree differential). All pass. That sweep also confirms the earlier judgement call: the net change to execute-phase.md is a SHRINK, and the size ratchet only gates growth, so reverting the append to the shared 2930-*.json ack fragment was correct — an ack would have been both unnecessary and factually wrong. Refs #1953 * chore(#1953): backfill changeset pr number to 3261 * docs(#1953): add the missing how-to for acting on a refactor proposal Reference and explanation shipped (COMMANDS.md, CONFIGURATION.md, FEATURES.md 159, ADR-1953) but the Diataxis how-to quadrant did not, and that is the one a user reaches for. CONTRIBUTING's required-docs table is 'new command -> COMMANDS.md + FEATURES.md', so CI was green on a gap. Enabling this feature is genuinely multi-step and no single page walked it: turn it on, tune the threshold, understand advisory vs strict, discover that strict needs a SECOND toggle on a DIFFERENT capability, and know what to do when a proposal appears. The two-toggle subtlety in particular was a footnote in a config table; here it is a section with both commands. Follows the shape of its closest siblings, resolve-edge-coverage-findings and resolve-prohibition-findings — both 'the loop surfaced a finding, here is what to do with it'. Includes a reason-code table for the silent cases, since the analyzer is deliberately quiet in six situations and a user who expected a proposal needs to tell 'nothing to report' from 'could not look'. Indexed from docs/README.md beside the other loop how-tos. Docs-only: exempt from the push gate, no re-verification, pass marker on 2af188b4 untouched. Refs #1953 * feat(#1953): warn when strict mode is on but nothing will actually block Closes acceptance criterion 5, which I had wrongly marked satisfied. refactor.trigger_strict records an untriaged proposal as an open deviation window, but a ship only STOPS if workflow.windows_enforce is also on — a toggle owned by the broken-windows capability that this feature neither sets nor requires. So a user could enable strict, believe ship was gated, and find out otherwise at ship time. The split itself stays: requires:["broken-windows"] would force-install the ledger on advisory users who never enable strict, and a ship:pre gate of our own would never fire because ship.md has no generic ship:pre gate dispatch. What was missing was discoverability, so that is what this fixes. `refactor evaluate` now emits a typed REFACTOR_STRICT_NOT_ENFORCING warning, naming the exact remediation command, whenever strict is on and either workflow.windows_enforce is off or broken-windows is unavailable. It fires only on a run that produced a candidate — with nothing to block on there is nothing to warn about, and warning every run would be noise. Reads workflow.windows_enforce through the same resolveConfigKey walk the router already uses for its own keys rather than a second config reader. Four tests cover the matrix: strict+enforce-off warns, strict+enforce-on does not, strict+ledger-absent warns, strict-off never warns. Also corrects a user-facing message in this same file that told the user to run `gsd-tools config-set` — the wrong form. docs/CONFIGURATION.md and the broken-windows capability both use `gsd config-set`, and gsd-tools is invoked as `node gsd-tools.cjs`, so the bare form may not resolve. The two adjacent messages in this file now agree. Refs #1953 --------- Co-authored-by: sim <sim@local>	2026-08-09 19:52:47 -04:00
Tom Boucher	0c413bbc9c	chore(#3059): close the ESLint glob-coverage escape and guard it (#3277) ... * chore(#3059): close the ESLint glob-coverage escape and guard it 62 tracked source files matched no `files:` glob in eslint.config.mjs, so ESLint skipped them entirely while `eslint .` still exited 0 — including all 26 files under hooks/, the enforcement machinery itself. Covers 56 of them (eslint-rules/, hooks/, bin/lib/, pi/, examples/, vscode/, the plugin shims, root *.mjs) and allowlists the 6 deliberate must-not-compile brand-typing fixtures with a recorded reason each. hooks/** is covered with n/no-process-exit deliberately off: a hook's whole contract is its exit code, several exits are load-bearing stdin-timeout guards where nothing else terminates the process, and ADR-0012/0174 scope the no-process-exit convention to the Command Routing Hub. bin/lib/ui-safety-gate.cjs is dual-mode, so it keeps the rule live and takes two targeted disables in its require.main===module tail instead. Adds scripts/lint-eslint-glob-coverage.cjs + a node:test drift guard so the class cannot regrow: allowlist entries require a non-empty reason, the list ratchets down only (a stale entry fails), and a tracked-count floor means a broken `git ls-files` fails rather than reporting clean. Closes #3059 * chore(#3059): apply review findings — correct the changeset count, add parser properties Isolated adversarial review, confirmed by rebuilding a byte-for-byte replica of the pre-change eslint.config.mjs: the changeset claimed 44 previously- unlinted files. The real figure is 56 (56 covered + 6 allowlisted = 62). That was user-facing CHANGELOG text and was wrong; corrected, along with three consequential figures in the design record. CLAUDE.md requires a fast-check property test for parsers and budget limits, and listTrackedSourceFiles is a parser. The standards review called this "satisfied in spirit"; it is not. Adds three properties driving the real exported parser through an injected execFile: extension totality/soundness including a trailing terminator, backslash-normalization totality, and CRLF/LF equivalence — the invariant the repo's recurring CRLF defect class breaks. Also de-duplicates the anchor rows onto one shared resolver, kept deliberately independent of the guard's own resolveFileCoverage so an anchor still fails if that resolution regresses, and records in the guard's header why the bin/install.js family is NOT allowlisted: it resolves to 2 rules under ADR-1703, so an entry would trip the allowlist_stale ratchet. * fix(#3059): make the coverage guard's git call container-safe The remote runner reported the guard degrading to `git_failed` on both Node lanes: fatal: detected dubious ownership in repository at '/work' The runner executes in a container where the repo is owned by a different UID, so git refuses to operate on it. The guard's degraded-verdict path worked exactly as designed — it reported the failure instead of throwing or falsely reporting clean — but a guard that cannot run in CI is not a gate. `git ls-files` is now invoked as `git -c safe.directory=* ls-files`. `-c` scopes the override to the single invocation and mutates no config file, and the wildcard is appropriate because this command only enumerates tracked paths in the repository it is already executing inside. Adds a regression test that captures the argv through the injected execFile seam and asserts `-c safe.directory=*` precedes `ls-files`, so the container case is pinned behaviorally rather than by reading the script's source. * chore(#3059): backfill changeset PR number pr:0 placeholder replaced with the real number now that #3277 exists. --------- Co-authored-by: sim <sim@local>	2026-08-09 19:35:34 -04:00
Tom Boucher	653f95e39f	chore(#2801): remove the hostBehaviors.reviewerCli deprecated alias (#3272) ... * test(#2801): failing-first suite for the hostBehaviors.reviewerCli alias removal Inverts the Phase 5a rows that assert the derived legacy alias still contributes a reviewer slug, and adds the removal-warning coverage the alias's exit needs (ADR-2782 D9). RED against unmodified production code, by design: the six shipped manifests still declare the key and collectReviewerWarnings emits nothing for hostBehaviors. Refs #2801 * chore(#2801): remove the hostBehaviors.reviewerCli deprecated alias ADR-2782 D9, Phase 7 — the final phase of epic #2782. The derived legacy alias survived one release (Phase 5a shipped in 1.9.0; 1.9.1 and 1.10.0 have since gone out), so it goes. A declared reviewer body is now the only route onto the reviewer roster. - deriveReviewerSlugs no longer reads runtime.hostBehaviors.reviewerCli - the key is stripped from the six manifests that carried it; each already declares a reviewer body whose slug equals its capability id, so the derived roster is unchanged at the same twelve slugs - collectReviewerWarnings emits a presence-based, non-fatal removal notice for any manifest still declaring the key, reaching both the build-time registry generation and the third-party overlay load path. The check runs before the reviewer-body early-return, because the manifest it exists for is the alias-only one that has no body. - hostBehaviors stays an open, unvalidated bag for its other 59 keys; this adds one keyed removal notice, not general validation Refs #2801 * refactor(#2801): give the reviewer-warning channel a typed IR Review finding: the new tests asserted with String#includes() on the warning prose, which CONTRIBUTING.md's 'Prohibited: Raw Text Matching on Test Outputs' bans in favor of a typed intermediate representation. Adds the IR beside the renderer rather than replacing it, which is the shape that section prescribes and bin/verify-reapply-patches.cjs already models: - REVIEWER_WARNING, a frozen code enum - REMOVED_REVIEWER_CLI_FIELD, so the emitting site and its test share one symbol instead of duplicating a literal - collectReviewerWarningRecords(cap), returning typed records collectReviewerWarnings(cap) keeps its exact string[] contract as a thin map over the records, so both production consumers are untouched. Every section-K row now asserts on record.code/field/capId and none on the rendered message. Locks the code surface, asserts the renderer stays one-to-one with the records, and migrates the pre-existing Phase 2 test on the same channel off prose matching. Refs #2801 * test(#2801): invert the section F alias fall-through regression row Caught by the remote runner: 2 unique failures on both Node lanes out of 31,692. tests/reviewer-lane-declarations.test.cjs section F — Phase 5a's isolated-security-review regressions — asserted that a blank reviewer.slug falls through to the hostBehaviors.reviewerCli alias rather than dropping the lane. That is the direct inverse of this phase's contract. The original rationale held only while the alias existed. With it gone there is nothing to fall through to: a blank body is not a declaration, and a declaration is the only route onto the roster. Inverted rather than deleted — the row carries the adversarial-review provenance for the slug trim, and removing a security regression guard to make a change pass is backwards. The duplicate row added earlier in section C is dropped instead; section F is its canonical home. Also corrects two count strings Phase 5b left at eleven while asserting twelve, which would misreport on failure. Refs #2801 * docs(#2801): give the removed reviewerCli flag a migration path The Reference edit alone satisfied CI — a file under docs/ moved, so lint-docs-required.cjs was green — while the task-oriented quadrant said nothing about the removal. A maintainer whose lane had just gone silent would have found the field documented as removed and no page telling them what to do about it. Adds a migration section to the how-to: the symptom, the verbatim warning they will see, the before/after manifest, and the note to keep the reviewer slug equal to the capability id so existing review.default_reviewers entries and --<slug> flags survive. Refs #2801 * chore(#2801): backfill changeset pr number to 3272 * feat(#2801): close the runtime.hostBehaviors vocabulary ADR-1016 closes twelve descriptor axes and rejects an open escape hatch in the descriptor. It never mentioned runtime.hostBehaviors, and that silence was read as permission: 59 keys across 18 manifests, 39 of them set by a single capability, validated by nothing. The reference docs went further and attributed the open seam to ADR-1016, which does not mention the field at all. KNOWN_HOST_BEHAVIORS enumerates the vocabulary. An undeclared key yields a non-fatal UNKNOWN_HOST_BEHAVIOR record on the same D4.3 channel as the alias removal notice, reaching both build-time generation and overlay install. Warning, never error, for the reason this phase exists: an error would hard-break an out-of-tree descriptor carrying a bespoke key with no deprecation window, which is what reviewerCli was given a release to avoid. Escalation is a separate decision. reviewerCli is excluded from the unknown-key sweep so it keeps its own notice with the migration pointer rather than drawing two records. A parity test binds the vocabulary to the shipped manifests in both directions, and a second asserts no shipped capability draws a notice, so the closure is provably inert in-tree. Records the decision and the miscitation as an ADR-1016 amendment. Refs #2801 * fix(#2801): bound and sanitize the unknown-key diagnostics Two findings from an isolated adversarial review of the closure commit, both proven by execution rather than asserted. MAJOR, introduced by the closure: the new Object.keys(hostBehaviors) sweep had no ceiling. An installed third-party manifest is bounded only by MANIFEST_MAX_BYTES, and an 8.69MB manifest with 800,000 keys produced 800,000 records and ~139MB of message text, retained for the registry's lifetime in OverlayMeta.diagnostics. Now capped at ten records plus a summary carrying omittedCount, mirroring capability-loader's existing slice(0,3) idiom. The same manifest now yields 11 records and 1748 chars. MINOR, newly reachable: manifest-supplied key names were interpolated raw. Unlike cap.id, which validateCapability gates on KEBAB_RE before these diagnostics run, hostBehaviors keys have no grammar check anywhere, so ANSI escapes and CRLF reached stderr and OverlayMeta.warnings intact. New describeKey replaces C0/C1 controls and clips at 80 chars. The file already had describeValue for this and applied it only to values. Both fixes land on the pre-existing reviewer.* sweep too — it carried the identical pair, and fixing only the new copy would leave the same defect one screen from its own fix. Refs #2801 --------- Co-authored-by: sim <sim@local>	2026-08-09 19:18:56 -04:00
Tom Boucher	58d73dd220	enhance(#3241): omit the codex per-agent model by default (#3276) ... * test(#3241): failing-first suite for the codex passive model posture Locks ADR-2313's D1-D5 before any production code exists, so the tests bind to the behavior rather than to whatever the implementation happens to do. Red-first (fail against the current tree): - the resolver path emits no `model` and no `model_reasoning_effort` - a whitespace-only model_overrides value yields no pin - isAnthropicFlavoredModel / CLAUDE_AGENT_ALIASES on model-catalog - the one-time install notice, and its once-per-install dedupe Regression guards (pass today, must keep passing): resolver-null via `inherit` and via absent runtime; a resolver that resolves to nothing; empty-string and non-string overrides; and the light-tier service_tier/model_verbosity fields, which are NOT coupled to the model pin and would silently regress if the implementation coupled them. Classifying each test as red-first or regression guard is deliberate. A test that passes on both sides of the change proves nothing, and this epic has already shipped two such rows before catching them. The whitespace case is a live defect, not a quirk: `' '` is truthy, survives the type guard, is not Anthropic-flavored, and is embedded verbatim as `model = " "` — the same class the #2310 guard exists to stop. Same function, same path, fixed in this phase per CLAUDE.md §3. Two matrix rows were dropped as vacuous rather than shipped green: a 64-char truncation case (the pinned notice interpolates no user-controlled value, so it cannot exhibit truncation) and a newline hazard that the input surface cannot reach. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * feat(#3241): omit the codex per-agent model by default Implements ADR-2313 D1-D5. generateCodexAgentToml no longer embeds the runtime resolver's per-tier Codex model, so an agent inherits the always-available session model instead of a pin a ChatGPT-account Codex may not expose. model_reasoning_effort disappears with it via the existing hasPinnedModel coupling (#838) — no logic change needed there. Supersedes #2517's embedding on the default path only. An explicit real-Codex model_overrides pin is still embedded verbatim, and the #2310 Anthropic-flavored guard is retained: the model_overrides route to it is still live even though the resolver route is now unreachable. The shared rule moves down a layer. CLAUDE_AGENT_ALIASES leaves model-resolver for model-catalog — a genuine leaf importing only node:path and its own JSON — with isAnthropicFlavoredModel defined beside it, and is re-exported from model-resolver so every existing importer is untouched. This is what lets Phase 2's install-check and Phase 3's sync consume the rule without taking the config-loader dependency model-resolver would have dragged into a module documented as pure read/verify with 33 dependents. A parity test fails if the two ever fork. Also fixes a live defect surfaced while writing the tests: a whitespace-only model_overrides value was truthy, survived the type guard, was not Anthropic-flavored, and so was embedded verbatim as `model = " "` — the same class the #2310 guard exists to stop, reached by a different route. Trimmed before the truthiness test. It is deliberately not routed through _warnCodexModelOverrideDropped, whose text would misdescribe a blank field as a mis-typed model. Adds the one-time install notice (maintainer direction, recorded as an ADR-2313 amendment): one stderr line naming model_overrides and the session model, deduped per install rather than per agent, and emitted only for the population that actually loses a pin. service_tier and model_verbosity stay decoupled from the model (#774); a regression guard asserts they still emit with nothing pinned. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(#3241): amend ADR-2313, add the model-catalog glossary entry ADR-2313 gains two dated amendments rather than edits to its merged text, since ADRs here are append-only. The first records that a deprecation notice IS offered, reversing the Migration section's "no deprecation window" position, and states why that position was wrong rather than just superseding it: the ADR identified the API-key population as losing something real and then declined to warn it, in the same document. Hyrum's guidance was applied to the recourse and not to the notice. The second records the whitespace-only model_overrides defect and notes that D2 always implied the fix — the implementation simply never enforced it and no test covered the case. CONTEXT.md gains a Model Catalog Module entry. The module had none, which is why the glossary gate passed without one: check-glossary-refs verifies that references resolve, not that modules are documented. The entry records why the Anthropic-flavored rule lives there rather than in model-resolver, so a later reader does not "helpfully" move it back. The Model Resolver entry is updated to point at its new home and note the back-compat re-export. docs/CONFIGURATION.md carried a claim that is now false: that the resolved tier ID is embedded into agent frontmatter at install time on codex and opencode. Corrected to name codex as the exception, with the 400 symptom and the model_overrides recourse. Changeset leads with the user-visible change and the migration line rather than the implementation, per the ADR's Hyrum's-Law analysis. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3241): only notice a lost pin when one was actually embeddable Review finding from an isolated reviewer. The deprecation notice gated on whether the runtime resolver would have returned *any* model, but the question that matters is whether that model would have been *embedded*. Those differ. The #2310 safety gate already rejected an Anthropic- flavored model arriving from the resolver path before Phase 1 — so for a mixed-runtime config resolving to a claude-* id against a Codex install target, the user never had that pin. The notice told them they lost something they never got, and pointed them at model_overrides for no reason. The existing #2310 test drives exactly that path but asserts only the emitted `model` line, never stderr, which is why it slipped through. Now covered. Deliberately unchanged: an Anthropic-flavored model_overrides value plus a legal resolver model fires BOTH the override warning and the notice. That is correct — pre-Phase-1 the guard dropped the override, execution fell through to the resolver, and the resolver's model was embedded, so that user did lose a pin. Two messages, two distinct true facts, and the prefixes differ (`gsd: warning — ` vs `gsd: notice — `) so the one-notice-per-install contract holds. A regression test now pins that behavior so it does not get "simplified" away. Of the three tests added, only the first is red-first; the other two pass on both sides by design and are labelled as guards — one against over-correcting the fix into silence, one against removing the intentional double message. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3241): reset the notice dedupe via a seam, not a require.cache bust The remote runner caught a regression I introduced: the #2760 post-write-validation test began failing with the validator override no longer intercepting. Cause, confirmed by trace rather than guessed: the new #3241 review tests deleted require.cache for bin/install.js and re-required it mid suite, to clear the notice's module-level dedupe flag. But runCodexInstall destructures `install` at file load, closing over the ORIGINAL module's exports. After the cache bust a second instance existed, so the test's `installModule.__codexSchemaValidator = ...` mutated the new object while the code under test still called the old one. The override silently stopped intercepting, the real validator ran and passed on GSD-emitted output, and the abort-and-restore path was never exercised. Cache-busting a module mid-suite breaks every later test that assumes a single instance, which every other test in the file is entitled to. So the fix is a seam, not a workaround: bin/install.js exports _resetCodexNoticeDedupeForTests(), and the three tests call it directly instead of reloading the module. The flag is module-level by design — the dedupe is per-install and install() already resets it — so a unit test driving generateCodexAgentToml directly needs an explicit way to reset it. That is now what it has. Swept the rest of the #3241 diff for the same hazard; this flag was the only shared module-level state introduced. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3241): reset both codex dedupe stores, not just the notice flag Second incomplete fix, same class one layer down. bin/install.js keeps TWO module-level dedupe stores and the require.cache bust I removed had been papering over both; my replacement seam cleared only one. _codexModelOverrideDroppedWarned is a Set keyed `${agent}::${value}`. tests/codex-config.test.cjs:558 already emits for `gsd-executor::sonnet`, so by the time the review test using the same agent and value ran, _warnCodexModelOverrideDropped was a silent no-op and the expected warning never appeared. The seam now clears both stores and is renamed to say so. Its comment records that per-install dedupe lives in module scope deliberately and that this is the single sanctioned way for a unit test to clear it. Swept bin/install.js for every other module-scope mutable a test could latch. Two are inert (capability registries assigned once at require time; selectedRuntimes computed once from argv). One is a genuine latent hazard and is deliberately NOT folded in: attributionCache (:1654) memoizes getCommitAttribution by runtime name for process lifetime, so two in-process installs of one runtime with differing attribution config would collide. It is unreachable from any current test and is a different concern from Codex warning dedupe, so it stays out of this PR rather than widening it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(#3241): correct the codex tier-routing how-to The docs gate forced the task-oriented quadrant and found the worst defect in this change's documentation surface. docs/how-to/configure-model-profiles.md carried a section titled "If you want tiered models on Codex" telling users to set runtime:codex + model_profile:balanced, promising "GSD resolves each tier alias to the Codex-native model and reasoning effort defined in the runtime tier map." That is exactly the behavior this PR removes — a how-to page confidently instructing users to do something that no longer works, which is worse than a missing page because it fails at the moment of use. Rewritten to state that Codex does no tier routing, give the model_overrides pin as the supported alternative, and name the two constraints on what may be pinned: it must be a real Codex model id, and the account must actually expose it — GSD cannot verify the second, so the honest advice when unsure is to omit the pin. Carries an upgrade note for both account types, since the change is a no-op for ChatGPT accounts and a real loss for API-key ones. Also tightened the same page's claim that Codex "embeds the resolved model" at install time — now true only of an explicit override. The re-install instruction it supports is still correct and still needed, so only the premise moved. Both the required-docs set (COMMANDS.md + FEATURES.md) and lint-docs-required.cjs would have passed before this commit, since CONFIGURATION.md and the ADR had already moved. Neither checks the quadrant a user in trouble actually opens. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#3241): backfill changeset pr number (#3276) --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-09 19:16:04 -04:00
Tom Boucher	a5706bd39d	enhance(#2596): validate a wave branch's committed diff stays in its declared scope (#3264) ... * test(#2596): failing-first suite for worktree-wave scope conformance Binds the advisory diff-vs-declared-scope check to behavior before it exists: the pure coverage predicate, the SUMMARY-artifact exemption and its parity with the rescue walker, the gauntlet integration (never flips ok, degrades on a git failure, survives a later block), the manifest normalizer's files_modified handling, and the --files negative-input matrix on record-agent/create. Refs #2596 * enhance(#2596): warn when a wave branch commits outside its declared scope The worktree-wave merge gauntlet validated branch, base, deletions, SUMMARY rescue and a clean worktree, but never compared a plan branch's actual committed diff against the files_modified the plan declared — so an executor that committed outside its brief merged into shared phase state silently. Adds an advisory scope-conformance check: when the manifest entry carries a declared scope, the gauntlet diffs HEAD...<branch> and appends one structured warning per path outside it. It never flips ok and never blocks the merge; promotion to a hard gate is a separate, disclosed change. With no declared scope no git subprocess is spent at all. Refs #2596 * docs(#2596): document the advisory worktree-wave scope-conformance check Records the optional --files flag on worktree record-agent/create, the advisory warnings channel cleanup-wave now emits, and its two deliberate noise limits (SUMMARY-artifact exemption, literal-prefix glob matching). Wires execute-phase to pass the plan's already-parsed PLAN_FILES. Refs #2596 * fix(#2596): close review findings on the scope-conformance advisory - share one path normalizer between the SUMMARY-artifact predicate and the scope comparison so the exemption and the check cannot drift - wire --files into the orchestrator-worktree dispatch, which created a worktree but never declared its scope, so the advisory silently did not apply on that backend; ADR-1239 requires both adapters share one check - correct the now-false blockquote claiming the check does not exist yet - add the fast-check property tests the repo requires for parser logic - add the record-agent/create parity test that Generative Fix Divergence requires for two surfaces implementing one rule Refs #2596 * fix(#2596): keep execute-phase.md under the frozen pre-phase-6 byte ceiling The one-sentence note added with the --files flag pushed execute-phase.md to 93708 bytes, past the ADR-857 PRE_PHASE6 cap of 93600 — the tightest of the three workflow size gates, and a hard cap an acknowledgment cannot clear. It failed three tests plus the differential attribution check. Condense the note to a one-line pointer (93543, 57 B of headroom); the full explanation already lives in docs/CLI-TOOLS.md and the dispatch step. The flag itself stays in the command, because the orchestrator reads this workflow at runtime and cannot pick it up from docs/. Acknowledge the remaining 143 B of growth by appending to the existing execute-phase.md fragment rather than adding a second one — the ack lint rejects two sources naming the same path. Refs #2596 * fix(#2596): make the execute-phase.md edit net-negative, not merely under the cap The size gate on this file is two assertions, not one: bytes < 93600 AND bytes <= 93400. The base is exactly 93400, so the file is at its budget and any growth trips the margin assertion — the previous fix cleared the ceiling but not that. Move the --files explanation to per-plan-worktree-gate.md, which already owns PLAN_FILES and carries no cap, and reclaim the rest from two clauses in the sentence being edited: the cleanup-wave rules phrasing, and a 'non-zero exit' the very next sentence already states. execute-phase.md ends at 93392, eight bytes below base. The flag itself stays in the command — the orchestrator reads this workflow at runtime and cannot pick it up from docs/. With no growth left, the acknowledgment is unnecessary and its byte delta was no longer true, so the shared ack fragment is restored byte-identical to base. Refs #2596 * docs(#2596): add the how-to for interpreting scope-conformance warnings The docs for this change were entirely Reference — the flag and the warning codes — with the task-oriented quadrant empty. Adds the page that answers the question an operator actually has when the advisory fires: what the two codes mean, that nothing is blocked so there is no failure to hunt for, how to tell whether the executor over-reached or the plan under-declared, and the three ways the check legitimately stays silent so an absence of warnings is not mistaken for proof of conformance. Refs #2596 * chore(#2596): backfill changeset pr number to 3264 --------- Co-authored-by: sim <sim@local>	2026-08-09 16:12:57 -04:00
Tom Boucher	f96cb44f85	enhance(#3248): disclose capability skills as an instruction surface (#3253) ... * test(#3248): failing-first suite for instruction-surface disclosure 28 matrix rows from 50-test-matrix.md. Rows requiring the new Disclosure.instructionSurfaces field fail today; rows 18-20/23-25 (the ADR-2363 D4 signature invariants) pass today by construction because the current code never reads skills/agents at all, and stand as regression guards for the implementation commit. Refs #3248 * feat(#3248): disclose capability skills and agents as an instruction surface ADR-2363 D5. A capability whose only contribution was skills disclosed nothing at install: summarizeDisclosure early-returned "ships no executable surfaces (declarative only)" because hasExecutable was false, while each SKILL.md body landed verbatim in the agent's instruction context. discloseExecutableSurfaces gains a fifth, NON-executable class, instructionSurfaces, collecting declared skills/agents stems through the same safeCollect wrapper as the four existing collectors, so a hostile value degrades only this class and the function stays total for any manifest shape. Nothing existing is edited: the collectors, hasExecutable, disclosureSignature and missingArtifacts are untouched. get_impact rates the symbol CRITICAL at 196 affected, which is why the design is strictly additive. D4 is implemented by omission and pinned rather than left incidental: adding, changing or removing skills/agents leaves disclosureSignature byte-identical, so no stored consent record is perturbed and no spurious re-consent fires. ADR-2782's conditional-append trick is deliberately NOT reused - it worked because no manifest could declare a reviewer body before that class existed, whereas skills predate this one, so a conditional append would re-sign every already-consented skill-bearing capability. The renderer is extracted as summarizeInstructionSurfaces and called from BOTH branches of summarizeDisclosure. Appending only at the end would never render for skill-only capabilities - the ones that need it - since those take the early return. That branch's "declarative only" claim is now conditional on there being no instruction surface either. The renderer iterates rather than spreading into push, so an unbounded stem count cannot throw RangeError, and tolerates the bare {} the CLI edge passes via `res.disclosure || {}`. Scope note: #3248's prose says "skill stems"; ADR-2363 D3 classifies instruction surfaces as "skills, agents". Shipping skills alone would leave an ADR deliverable owned by no phase, and the epic has no Phase 2. Agents are the same shape at no extra cost. Narrowing back is a two-line change. Ratifies ADR-2363 (Proposed -> Accepted) and adds the owed ADR-1244 back-link. Closes #3248 * fix(#3248): escape consent-prompt values and narrow disclosure to skills Two review findings, both of which made the previous commit wrong. BLOCKER (isolated adversarial review). Every manifest-supplied value interpolated into a consent-prompt line was rendered unescaped. Those lines are joined with \n and written RAW to stderr on the needs-consent path (capability-command-router -> cli-exit runMain), so a stem carrying a newline forged additional lines indistinguishable from genuine GSD disclosure text, and an ANSI escape could clear or rewrite lines already printed. That defeats the informed-consent guarantee this change exists to provide, and is a prompt-injection vector against any agent that reads the stderr text to decide whether to retry with --yes. The hole was not unique to the new class - hook event/script, command family/module/router, every MCP field, and every reviewer-lane field were equally unescaped. Fixing only the new one would have created the generative-fix divergence this repo tracks, so renderValueForPrompt is applied to all five classes through one helper, guarded by a parity test that fails if a future class skips it. Escaping is identity for ordinary names, so no well-formed manifest's output changes. The disclosure OBJECT stays verbatim - only the rendered LINE is escaped - because the signature and every consumer reasoning about identity depend on the declared value. NARROWED to skills only. The previous commit also collected agents, arguing ADR-2363 D3 classifies instruction surfaces as "skills, agents". Verified against staging: stageSkillsForRuntimeAsSkills takes a registry and unions third-party skills in via readInstalledCapabilitySkill, while stageAgentsForRuntimeWithConverter takes only a source directory and has no registry-aware path. Third-party agents are never staged into the instruction context, so disclosing them would have put a false claim in a security prompt - worse than the scope creep two reviewers flagged it as. D3's classification stands; D5 now records that Phase 1 implements the skills half and that whether agents should be staged at all is an open maintainer question. Also reverts the premature ADR-2363 ratification. The previous commit flipped it to Accepted and asserted "#3248 merged" while this branch IS #3248 and is unmerged. Status returns to Proposed, and the ADR-1244 back-link - owed only on ratification - is withdrawn. Adds the fast-check property suite CLAUDE.md requires and the direct precedent (reviewer-trust-disclosure) already had: totality, D4 signature invariance, D3 hasExecutable invariance, and renderer totality over adversarial manifests. Refs #3248 * chore(#3248): correct changeset scope claim and backfill pr number The fragment was written against the pre-narrowing commit and still advertised 'skills and agents'. 4d26887e narrowed disclosure to skills only - third-party agents are never staged into the instruction context - but did not touch the fragment, so the release notes would have carried a claim the code does not implement. Also backfills pr:0 -> 3253 and names the prompt-escaping fix, which is user-visible and was absent from the original body. Changeset-only; no code or test changed, so the gsd-test pass recorded for 4d26887e still describes this tree's behavior. Refs #3248 --------- Co-authored-by: sim <sim@local>	2026-08-09 13:52:29 -04:00
Tom Boucher	bc5619dd27	docs(#3247): record the capability instruction-surface trust model (#3249) ... * docs(#3247): record the capability instruction-surface trust model ADR-2363 records the trust posture for third-party capability SKILL.md bodies, which #2322/#2340 made agent-invocable without any content-level control. The path-level protections that fix shipped are all present; no content scanner exists, and external-descriptor-trust.cts never had one. Nothing was bypassed - the control did not exist and the boundary was never written down. D1 records the posture: skill bodies are trusted, unscanned agent instructions. D2 rejects content scanning on Kerckhoffs (a shipped rule set is readable by the adversary who installs it), on threat-model non-transfer from ADR-1577 (there, instructions are anomalous inside data; here they are the payload's legitimate form), and on Goodhart (a scanned-OK line displaces the judgment the consent prompt exists to provoke). D3 replaces the executable/non-executable binary with three classes, adding instruction surface. D4 keeps instruction surfaces out of the v1 disclosureSignature. The signature is NOT the activation binding - hasProjectConsent compares contentHash only, and a global install carries no consent record at all. What re-encoding would do is perturb the signature of every skill-bearing capability and fire a spurious re-consent prompt on its next upgrade, which is what ADR-2782 D4 rule 5 already forbids. If instruction surfaces ever need to be signature-bound, that lands as a versioned v2 signature with a migration, never an in-place re-encoding. Corrects capability-trust-model.md, which claimed skills get lighter consent because they do not execute code - true, and not the relevant property, since the agent is the interpreter. Adds the author-side boundary to develop-a-capability.md and links it from publish-a-capability.md. Both state that per-skill disclosure at the consent prompt lands with #3248 and does not happen today. Docs-only. No behavior change; no consent record perturbed. D5's mechanism is Phase 1 (#3248), which is why the ADR is Proposed. Refs #2363 * chore(#3247): backfill changeset pr number to 3249 --------- Co-authored-by: sim <sim@local>	2026-08-09 11:47:54 -04:00
Tom Boucher	27aa40f65e	fix(#3023): stage pi's shared hook bundle outside pi's reserved hooks/ directory (#3175) ... * test(#3023): failing-first guard — pi must not stage hooks in its reserved dir pi reserves <configDir>/hooks as its deprecated extension location and warns on every startup when it exists. Assert a pi install stages the shared hook bundle under gsd-hooks/ instead, manifests it there, and never creates hooks/. Also adds pi to the local-scope dir table in install-shared.cjs: pi was in RUNTIME_META but not LOCAL_DIR_NAME, so scope:'local' resolved path.join(root, undefined) and no local pi install could be exercised. Fails before the fix. Verified via the remote runner. * fix(#3023): stage pi's shared hook bundle outside pi's reserved hooks/ dir pi reserves <configDir>/hooks as its now-deprecated extension location and warns on every startup when that directory merely exists — checkDeprecatedExtensionDirs() guards the warning with a bare existsSync(), unlike its tools/ sibling. GSD staged its shared hook bundle exactly there, and pi's advised remediation (move it to extensions/) would break the adapter's paths and expose GSD's .js helpers to pi's extension auto-discovery. The bundle directory name is now runtime-descriptor-driven: hostBehaviors .sharedHooksDirName, defaulting to 'hooks' so all 18 other runtimes are byte-identical. pi sets 'gsd-hooks'. The name is validated as a single path segment — separators, dot-only segments, trailing dots, absolute paths, NUL, and Windows reserved device names all fall back to the default, because the value is joined onto a user's config root and written to. Renamed in place rather than relocated: hook scripts resolve siblings via __dirname/.., so a depth change would silently break them. - install / uninstall / manifest sites all read the resolved name - pi/gsd.cjs probes gsd-hooks then hooks, so dev checkouts and half-upgraded trees still resolve; the never-throws contract is preserved - new migration 009 retires the legacy pi hooks/ dir on upgrade, using a new non-recursive remove-empty-dir engine primitive (rmdirSync only, symlink-refusing, containment-guarded); ADR-0008 amended accordingly - fixes two latent name-dependencies the rename exposed: the stale-hook scan and the injection scanner's self-exclusion both hardcoded 'hooks' Verified on the remote runner. Closes #3023 * fix(#3023): close review findings and align emitted provenance with the rename Adversarial review found two defects, and the remote runner found four failure clusters. All fixed here. Review BLOCKER — detect-custom-files was blind to the renamed bundle. GSD_PREFIX_MANAGED_DIRS in gsd-tools.cjs hardcoded 'hooks', so for pi the whole gsd-hooks/ tree was invisible to the custom-file scan and user-added files there were never backed up before the next update's clean-install wipe. The dir set now resolves via the .gsd-runtime marker plus the shipped capability registry (never bin/install.js, which is not shipped into installed trees), and falls back to scanning every known candidate when the runtime cannot be determined — over-scanning is safe, under-scanning is the data loss. Review MAJOR — the pi adapter bound to an empty bundle. resolveSharedHooksDir accepted any directory, so an interrupted install left gsd-hooks/ winning over a fully-staged legacy hooks/ and every hook silently no-opped. A candidate now qualifies only if it is non-empty. Remote-runner clusters: - emitted-provenance had no rule for the gsd-hooks/ family; added two pi-scoped rules pointing at the same sources the existing hooks/ rules use. The table is total, so an unattributed family is a hard failure by design. - pi tests in install-minimal-hooks and the install integration suite asserted the old layout; updated to derive the dir name from the descriptor rather than hardcoding either name. - 19 unrelated-looking failures on node22 only were a leaked fs mock: t.after() runs in registration order, cleanup was registered before mock.restoreAll(), and node22's JS rimraf calls the public fs.rmdirSync while node24's native path does not — so the EACCES stub leaked process-wide on one lane. Restore now runs first. Verified on the remote runner. * fix(#3023): honor PI_CODING_AGENT_DIR, ack the rename ripple, fix expandTilde pi resolves its agent dir as PI_CODING_AGENT_DIR ?? ~/<CONFIG_DIR_NAME>/agent (packages/coding-agent/src/config.ts). GSD's pi descriptor declared an empty configHome.env, so a user with that variable set had GSD installed where pi never looks. Added the env name; the dot-home-nested resolver already handled the override, so no resolver logic changed. Also fixes expandTilde in the shared runtime-homes resolver, found while adding that: it hardcoded os.homedir() and ignored the opts.home every caller threads, so EVERY runtime's tilde-valued env override (claude, antigravity, windsurf, pi) silently resolved against the real home. That is a correctness bug and a test-escape hazard — a sandboxed test asserting on a tilde override reached the developer's actual home directory. Now threaded through every branch; behavior with no injected home is unchanged. Adds the emitted-drift ack fragment for the 58 pi paths whose emitted location moved with the rename. The provenance rules satisfy the totality gate; the differential gate needs the ack because the hook sources are byte-unchanged — only the installer's target directory moved. The two hook files this branch genuinely edits stay attributed and are not double-acked. Note on piConfig.configDir: it is read from pi's OWN installed package.json (getPackageDir walks up from pi's __dirname), alongside piConfig.name — a white-label setting for a redistributed pi fork, not a per-project user setting. Documented accordingly rather than treated as an unsupported override. Verified on the remote runner. * fix(#3023): reject blank env overrides, pin adapter/descriptor parity Three review findings, all fixed. A whitespace-only config-dir override was accepted verbatim: the guard was `if (val)`, falsy only for the empty string, so PI_CODING_AGENT_DIR=' ' resolved to a literal three-space directory name instead of falling back to the descriptor default. Fixed across every env-consuming branch — dot-home, dot-home-nested, all three xdg steps, and generic-agents-root — not just pi's. Non-blank values are still never trimmed, so '~/My Agent Dir' keeps working. pi/gsd.cjs's probe list and the descriptor were two independent sources of truth for the bundle directory name; a future rename would have desynced them silently and left every pi hook quiet with no error. The probe list stays deliberate — it must resolve in a dev checkout and a half-upgraded tree, where the registry's answer would be wrong — so this adds the parity assertion the repo's generative-fix-divergence rule calls for: the descriptor value must be the FIRST candidate, and the default must remain present. Changeset body rewritten to cover the two later user-facing fixes it had not caught up with. Verified on the remote runner. * chore(#3023): backfill changeset PR number * fix(#3023): anchor injection-scan patterns and fix a macOS detection hole CI's security job flagged CONTEXT.md:124 — pre-existing prose reading 'not the same fact as a genuinely empty or absent one'. The match was the 'act as a' INSIDE 'f-act as a': the pattern had no left word boundary, so any word ending in act tripped it (fact, impact, contract, artifact, interact, redact, abstract). My four-line CONTEXT.md edit dragged the latent false positive into this PR because the scan is diff-scoped by file but reads whole files. Anchored with (^|[^[:alnum:]]) rather than rewording maintainer-owned prose, which would have left the class alive for the next PR touching any file saying 'fact as a'. Auditing the rest of the list for the same class surfaced a real detection hole: the eval/exec/Function patterns matched a quote via \x27, a GNU-grep-only hex escape. BSD/macOS grep reads it as four literal characters, so single-quoted eval('...')/exec('...') payloads were NEVER detected there while passing on GNU-grep CI. Replaced with a literal apostrophe class. Boundaries were added only where a real word-suffix collision exists; exec, jailbreak, developer mode and the role-manipulation family were audited and deliberately left unanchored. 22 new cases cover both directions — the false positives now scan clean, and every real payload still fires, including the quote/punctuation/start-of-line boundary forms. Also builds this branch's injection test fixture at runtime instead of carrying the literal phrase, so the payload keeps its teeth without tripping the scan. Verified on the remote runner. --------- Co-authored-by: sim <sim@local>	2026-08-07 13:41:21 -04:00
Tom Boucher	0e6fa2e2cf	enhance(#3118): close the dead injectables and the shell projection follow-on — Wave 4 (#3124) ... * test(#3118): failing-first coverage for the dead injectables and the shell projection Adds the counter-tests Wave 4 closes against, before any fix: - antigravityWatermark had zero test references. The four existing tests that look like watermark coverage hand the fallback a literal mark and never call the producer, so nothing pinned whether a real run's mark is correct. Covers all six branches plus the non-object cache classes. - Pins the fail-open: a transcript read that throws reports lines:0, indistinguishable from a genuinely empty transcript, and the consumer then replays a previous run's review as this run's. - Pins the export-line escaping across the repair, persist and win32 bash lanes, including the parity assertion that they must not diverge. - sliceCurrentPositionSection: empty-vs-absent, fenced heading, second occurrence, H3, CRLF. - Proves deps.progressProvider is inert by supplying a throwing stub to all ten transition intents. Verification through the remote runner only. Refs #3118 * fix(#3118): distinguish an unreadable transcript from an empty one antigravityWatermark's final read can throw on a transcript that indisputably exists. It returned lines:0, which is the same value a genuinely empty transcript produces, so the caller could not tell the two apart. antigravityTranscriptFallback derives its skip from that count. A mark of {convId:'c1', lines:0} for a conversation that pre-dates the run makes it skip nothing and return the last PLANNER_RESPONSE in a transcript written before this run started — a previous review presented as this one's, which is exactly what the function's own 'never stale' docstring promises cannot happen. The unreadable case now sets unreadable:true and the fallback declines for a same-conv-id unreadable mark. An absent or empty transcript is untouched: those genuinely have zero prior lines. * fix(#3118): escape the export line for the file it lands in, not the echo Three lanes emit export PATH="<dir>:$PATH". repair escaped it with escapePosixDoubleQuoted; persist and the win32 Git Bash lane escaped it with escapeSingleQuotedShellLiteral instead. The single-quoting is correct for the echo, so nothing runs when the user pastes the command. But the bytes appended to ~/.bashrc are the export line itself, and inside double quotes in an rc file a $(...) or a backtick in the directory name is command substitution that runs on every new shell. Those characters are legal in a path on both POSIX and Windows, so the path was reachable. projectPathExportLine is now the single source of that line and escapes for its final rc-file context; each lane still applies its own transport escaping on top. fish keeps the single-quote escaper — its value really does stay single-quoted. The cmd.exe lane interpolated into a cmd double-quoted string with no cmd-level escaping, so a quote closed the region and &cmd& ran. A quote is reserved on Windows and cannot appear in a real path, so there is no correct command to suggest: the win32 lanes now fail closed for one. Metacharacter-free paths render byte-identically on every lane. * fix(#3118): drop a stray carriage return and a deps field nobody reads locateCurrentPosition subtracted a fixed one byte to exclude the newline before the next heading, which assumes LF. On a CRLF document the slice kept an unpaired trailing carriage return. It now walks back over the newline and over a preceding carriage return if there is one. StateTransitionDeps also required a progressProvider that 33 sites supplied and no site ever called. A required field nothing reads widens the module's interface without changing its implementation, which is the shape epic #3051 cites as its reason for refusing blanket injection. Removed along with the ProgressRecord alias that existed only as its return type; state-document.cts's unrelated interface of the same name is untouched. * fix(#3118): stop an empty span duplicating bytes, and name the empty results Three findings from the isolated review pass. locateCurrentPosition could return end < start when the section was empty and the next heading followed with no blank line between. Every mutator splices with slice(0,start) + body + slice(end), so an inverted span duplicated the region between them — a blank line silently inserted into STATE.md on every transition, two bytes on CRLF. The span is now clamped, and an empty section is a zero-length span, which is what it always meant. The win32 fail-closed path left the installer printing 'Add it with one of:' with nothing under it. An empty shellActions folded two different facts together, so projectPathActionProjection now carries a frozen PATH_ACTION_REASON and the installer branches on it. Two empty results with different causes staying distinguishable is the subject of the epic this belongs to. fish_add_path parses a leading dash as an option, so a directory named -v printed 'No paths to add' instead of being added. Verified against fish 4.8.1: the end-of-options separator fixes it. Replaces the console-prose test the second fix first arrived with — a regex over captured stdout is what CONTRIBUTING prohibits, and the typed reason is the surface it asks for instead. * fix(#3118): escape TOML control characters, and stop a test name overstating Five findings from the two review axes. escapeTomlDoubleQuotedString escaped only backslash and quote. TOML basic strings also require U+0000-U+0008, U+000A-U+001F and U+007F to be escaped, so a value carrying a raw newline or NUL wrote a config.toml no parser accepts — rejecting the whole file, not just that value. Four of its call sites write real config. Tab stays raw; the grammar exempts it. The byte-identity test claimed every lane was unchanged for an ordinary path, which is false: fish now takes the end-of-options separator on every path, not only hostile ones. Renamed, and the one intended delta now has its own named test instead of hiding inside a claim that read as broader than it was. Also: exact-equality assertions in place of substring checks that could pass on a subtly wrong escape, newline and null-byte cases for all five quoting primitives, and a temp dir registered with t.after so it is removed when an assertion fails. * docs(#3118): add the changeset fragments * fix(#3118): degrade instead of throwing on a null conversation cache A cache file whose whole content is the literal null — what a truncated or zeroed write leaves behind — made both antigravityWatermark and antigravityTranscriptFallback throw. JSON.parse('null') succeeds, so the try/catch wrapping the parse never fired, and resolveConvId then called hasOwnProperty on null. Both functions advertise the opposite; the existing test next to them is named 'a missing cache or transcript degrades to empty, never throws'. Parsing successfully is not the same fact as the payload being usable, and a guard that only wraps the parse cannot tell them apart. resolveConvId is now total for any non-object input, so one guard covers both callers. Caught by the null case in this wave's own cache matrix. * test(#3118): correct a stale fish expectation and a parity comparison The pre-existing 'POSIX persist mode escapes single quotes' test pinned fish_add_path without the end-of-options separator this wave adds, so it asserted behavior that is no longer correct. A repo-wide scan found one such hardcoded expectation; every other site derives its expectation from the projection. The new parity test compared the token from a POSIX path against the win32 lane, which posix-normalizes its input first — two different inputs, so the tokens differed for a reason that had nothing to do with the parity it claims to check. It now derives the win32 expectation from the same input the lane receives. * docs(#3118): reword a comment the injection scanner reads as an instruction The scanner pattern act\s+as\s+(?:a|an|the)\s+ carries no word boundary, so 'the same fact as the payload' matched on the tail of 'fact'. Reworded per the documented remedy for this collision. The missing boundary is a scanner defect rather than a prose problem — any contributor writing 'fact as the' trips it — but the pattern is gate plumbing, which the sibling epic owns, so it is surfaced rather than changed here. * chore(#3118): backfill changeset pr number to 3124 * chore(#3118): backfill changeset pr number to 3124 * fix(#2784): make the negation scan single-pass and index it correctly Three defects in the negation suppression added by #3127, all in one block, none of which had a test. The pair scan was verbs.some(nouns.some(...)) with a slice and a split per pair, so it grew cubically with clause length: 1.1ms before that PR and 8462ms after, on 800 verb+noun pairs in one clause. api-coverage's property test generates documents large enough to reach the runner's 600s file cap, which is why it hangs as 'fail 0, cancelled 1' rather than failing an assertion. Every (verb, noun) window is a subset of the single widest one, so one scan of that window answers the same question in a linear pass. Verified equivalent against the old predicate over 20,000 generated clauses. Both checks also subtracted clause.start from offsets that collectTerm- Matches already returns clause-local. The first clause on a line has start 0 so it worked there and nowhere else: later clauses went negative, and slice reads a negative index from the end, so suppression silently examined unrelated text. The comment claimed 'without any API integration' was suppressed. It is not — the qualifier sits outside the two-word lookback and the noun precedes the verb. Widening the window would trade a false positive that costs one declaration line for a false negative that slips a real integration past a blocking gate, so the behavior stands and the comment now says so. Pinned by a test. The qualifier sets were also rebuilt for every line of every document.	2026-08-06 23:57:05 -04:00
Tom Boucher	7203011400	feat(#3072): ship the deferred MCP served catalog (resources + prompts) (#3083) ... * test(#3072): add failing-first coverage for the mcp served catalog 55 input-class rows from the phase test matrix, across four suites: the catalog module over injected readFile/readDir seams, the protocol surface through handleMessage, the install-vs-catalog parity gate, and fast-check properties for uri round-trip, traversal refusal, and pagination partition. src/mcp-catalog.cts lands as a skeleton whose functions throw, so the suites fail on BEHAVIOR rather than on a missing module. The REASON enum is real so tests assert typed codes instead of message prose. Hostile coverage for the one client-controlled path surface (resources/read): dot-dot and backslash traversal, percent- and double-encoded traversal, absolute posix and windows paths, file:// scheme, null byte, symlink escape, unindexed sibling, non-string and empty uri, wrong root segment. IO faults are injected by monkeypatching the seam, never chmod 0o000 - root bypasses mode bits, so a permission-based test silently passes with zero coverage in root CI. Refs #3072 * feat(#3072): serve the mcp catalog as resources and prompts gsd-mcp-server now serves GSD's own content alongside its three tools: the workflow, reference and command tree as MCP resources (resources/list, cursor paginated, and resources/read over gsd://<segment>/<relpath> uris) and the 71 commands/gsd/*.md as MCP prompts keyed by bare command name. initialize advertises resources and prompts, and deliberately does not advertise subscribe or listChanged - the catalog is fixed for a server process lifetime, so declaring a notification we never send would be a lie a host acts on. Composition scope is SHARED, not re-declared. shouldCompose lives in src/mcp-catalog.cts and bin/install.js now imports it instead of carrying its own regex, so the served catalog and the installed file floor cannot drift on what gets composed. Proven behavior-preserving across all 2871 tracked paths plus windows-backslash, absolute and near-miss-prefix cases: zero mismatches. tests/mcp-catalog-parity.test.cjs asserts served text equals the installer composition-stage text over the real tree, with anti-vacuity guards requiring both a marker-bearing workflow and a non-composed file in the comparison set. Two measurements corrected the literal issue text. Composition is scoped to gsd-core/workflows/ only, because a reference or command that documents marker syntax with an unfenced example would otherwise be parsed as carrying a real marker and have that line lossily dropped. And parity is asserted at the composition stage rather than against an emitted runtime tree, since install applies per-runtime path rewrites afterwards and the catalog is host-agnostic, so byte equality with any one runtime would be false by construction. resources/read is the one client-controlled path surface and is guarded in two independent layers: the uri must be an exact key in the prebuilt index, which defeats every traversal string by construction, and the mapped path is then re-checked with validatePath so a symlink planted inside a root after indexing is still refused. Also fixes a real drift defect found while here: SERVER_VERSION was hardcoded 1.7.0 while the package is at 1.9.1. It now resolves lazily from VERSION or package.json, reusing the precedent in runtime-artifact-conversion. Closes #3072 * test(#3072): make the catalog parity gate drive the real installer Review found the parity gate vacuous: it never imported or spawned bin/install.js, and recomputed the installer side with the SAME shouldCompose and composeWorkflow the catalog calls internally. It therefore proved only that src/mcp-catalog.cts is self-consistent. The old row 52 compared shouldCompose against a regex literal frozen in the test file rather than against the installer at all. An inline divergent regex re-added to bin/install.js - the exact regression ADR-1671 asks this gate to catch - would have left the suite green. The gate now spawns a real bin/install.js and compares the composition DECISION, observed as gsd:section marker survival, against what the catalog serves for the same files. Marker presence is the right observable because the installer applies per-runtime path rewrites after composing while the catalog applies none, so raw byte equality between the two surfaces is false by construction and must not be asserted. Sensitivity was proven, not assumed: overlaying the shouldCompose export that bin/install.js imports so it always returns false makes a real spawned install leave autonomous.md's markers in place while the catalog still strips them, and the row 48 assertion diverges. Anti-vacuity guards are kept and extended - the comparison set must be non-empty, must contain a workflow that actually carries markers, must contain a file the predicate declines to compose, and the install must have emitted a non-zero file count. The marker-documenting reference case has no instance in the real tree, so it uses an overlay fixture built with the same technique workflow-fragments-emission.install.test.cjs already uses. Renamed to .install.test.cjs so it lands in the install suite it now belongs to. Refs #3072 * test(#3072): retarget the unknown-method assertion off a now-implemented method tests/gsd-mcp-server.test.cjs used 'resources/read' as its example of an UNKNOWN JSON-RPC method. The served catalog implements that method, so it now returns -32602 (no uri supplied) rather than -32601. The remote runner caught it deterministically on both linux lanes: -32602 !== -32601. The test's intent is still correct and worth keeping, so it is corrected rather than deleted or weakened. It now uses 'resources/subscribe', which the server deliberately does not implement and deliberately does not advertise in initialize's capabilities, because it never sends the corresponding notification. That turns the assertion into a real contract - the advertised capability surface and the implemented method surface agree - instead of an arbitrary method name a future feature could invalidate the same way. Swept the rest of the suite for other assertions pinning the newly implemented methods; this was the only one. Refs #3072 * chore(#3072): backfill changeset PR number 3083 * test(#3072): make the catalog fake fs separator-agnostic for windows CI caught this on windows-latest (22 and 24): every catalog fixture indexed ZERO entries, surfaced by the anti-vacuity guards as 'fixture catalog must actually index resources for this property to mean anything'. Mechanism: makeFakeFs keyed its dirMap/fileMap on POSIX-joined paths (${root}/${rel}), while production buildCatalog looks paths up with path.join, which is backslash-separated on Windows. Every lookup missed, tryReadDir returned null, and the catalog came back empty. Production is NOT at fault and is unchanged. The same CI run proves it: on windows-latest the real-filesystem tests all passed, including 'installer composition decision matches the served catalog for every file in the real installed tree' and the row-51 non-vacuity proof against a real spawned installer. A real Windows fs accepts both separators; the FAKE did not, so the fake was the unfaithful one and is what changed. Lookup keys are now normalized unconditionally with .replace(/\\/g,'/') in readDir and readFile - never path.sep-conditional, never platform-gated. The row-42/43 injected-fault wrappers got the same treatment, since they compared raw production paths against POSIX-literal fixtures. No assertion was weakened, and the anti-vacuity guards that caught this are untouched - they are the reason this surfaced as a loud failure instead of a suite that silently asserted nothing on Windows. Refs #3072 --------- Co-authored-by: sim <sim@local>	2026-08-05 13:30:55 -04:00
Tom Boucher	8f75e27554	fix(#3045): fail closed when an executor dispatch drops its resolved isolation (#3069) ... * feat(#3045): deny an executor dispatch that drops its isolation flag Every isolation gate already resolved correctly. The resolved value then reached the executor through a prose instruction telling the model to substitute it into a call the model composes itself, and nothing verified the substitution. When it was dropped, the executor edited and committed in the user's primary checkout with no consent and no warning. A prose backstop would be the same class of artifact as the defect, so this is a shipped PreToolUse hook on the Agent tool. It fires at the instant of the call rather than being read once at the top of a workflow, which is the only placement the model cannot skip. The guard is inert unless it can positively establish that this is a GSD project, that the project resolves to harness isolation, and that the dispatch targets an executor. A non-GSD repo has no invariant to enforce. Where it cannot read the configuration at all, it denies rather than assuming, with its own reason -- a guard that cannot verify must not answer safe. A malformed payload allows rather than throwing. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * feat(#3045): extend the isolation guard to Cursor Cursor is the second of only two runtimes that resolve harness isolation, so shipping the guard for Claude alone left half the exposed surface unguarded while the changeset implied it was covered. The two runtimes fail differently. On Claude the harness flag is a per-dispatch kwarg the model must copy into a call it composes, and the defect is that it can be dropped. On Cursor the flag is --worktree, which applies to the whole session, and the subagent-start payload carries no isolation field at all. There is no flag to check, so the guard verifies the effective state instead: whether the workspace is genuinely running outside the user's primary checkout. That is a stronger check than the Claude one because it tests reality rather than intent, and it is commented so nobody later rewrites it into a flag check. Isolation is established two ways, either sufficient: the workspace resolves to a linked git worktree, or it sits under the worktree root Cursor manages. The second matters because a directory Cursor placed there is a legitimate isolated session even before it becomes a distinct git worktree, where linkage alone would report no repository. Detecting linkage required a new primitive rather than the existing context resolver. That resolver short-circuits on finding a local .planning directory before it ever compares the git directory to the common one -- and an isolation worktree normally has its own checked-out .planning. Reusing it would have read a correctly isolated session as unisolated and denied it, which is the failure direction that gets a guard switched off. The comparison is now its own shortcut-free function that the resolver delegates to after its own shortcut, so existing behavior is unchanged, and the case that would have broken is pinned. The subagent type is checked before any configuration is read, so an unreadable config cannot deny a dispatch this guard would never have enforced against. The input-schema comment on the Cursor hook documented only the fields common to every event and omitted the ones specific to this one. That omission cost a halt during this work; it now documents both. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3045): enforce the resolved dispatch decision, not the host capability The guard keyed on the registry's dispatch.isolation, which says only that a runtime is CAPABLE of harness worktrees. The decision that actually governs a dispatch is the one the workflow resolves after gating, and that legitimately comes out as sequential in three documented cases: a project setting use_worktrees false, a per-plan submodule intersection, and the base-check auto-degrade. The workflow tells the model to omit the flag in exactly those cases, and the guard was denying every one of them. The third case matters most. The preceding fix made the base-check degrade on git timeouts and a missing git binary, where it had previously answered "safe". That correction is right, and it means a transient hang now degrades to sequential far more often than before -- so the two changes composed into a trap where the workflow behaved exactly as designed and the guard blocked it. The workflow already resolves isolation in shell, deterministically, which is what makes it a trustworthy source in a way the model-authored call is not. It now records that resolved value through a dedicated verb, and both guards read it first. A fresh record is authoritative, so sequential dispatches pass untouched. Absent or stale, the guards fall back to the capability check combined with the project's use_worktrees setting, which still covers the case that never reaches the workflow. Also widened the matcher to accept Task alongside Agent, since a host that names the tool Task would otherwise leave the guard silently inert while implying coverage; stopped assuming Claude when no runtime is declared, which is the shipped default and would have demanded a Claude-only argument elsewhere; and made a non-git project inert rather than denied, since advising a worktree session is not actionable without a repository. The original diagnosis never modeled sequential mode as legitimate. That omission is what let this through, and it is now recorded there. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3045): record at resolution and bind the record to its dispatch Two independent reviews converged on the same failure: the guard was fail-open in a default install, so it did not catch the defect it exists to catch. A shipped project carries no runtime key, which made "runtime not confidently known" the common case rather than a corner one. A record asserting that isolation was required but carrying no flag then fell through to a capability lookup that answered "none", and the dispatch was allowed. The flag itself only arrived from a second shell block -- the same block a model dropping the argument would also skip. A test had pinned that behavior as intended. The record is now written by the resolver, as an unavoidable consequence of asking for the value, rather than by a step the model is told in prose to go and run. A guard against a prose-carried value cannot itself depend on prose. Mode, flag and identifiers are written together and atomically, so the flagless window is gone, and a record asserting isolation with no resolvable flag now denies instead of degrading. Runtime is also resolved from the installer's own recorded default, which makes confident resolution the normal case. The per-plan submodule gate degrades after the phase-level decision and never re-recorded, so a plan that legitimately ran sequentially was denied against a still-fresh phase record. It now records its own, scoped to the plan. A record also authorized any dispatch for four hours. One phase degrading to sequential could silently license an unisolated dispatch in the next. Records now carry phase and plan, the guards require them to match, and the window is minutes rather than hours -- the resolver rewrites it before every dispatch, so a long window bought nothing and only widened the hole. The flag validator rejected any value beginning with two dashes, which is exactly the form Cursor and Windsurf declare, so their real value could never have been stored. Writer and reader also derived the record path differently and diverged inside a linked worktree without local planning state. The predictable path remains a way to silence the control without leaving a trace in the diff. It grants no access an agent with shell does not already have, so it is documented as accepted rather than redesigned around. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3045): correct the staleness boundary and unmask a vacuous parity test The remote runner returned twenty failures. One was a real production defect the boundary case existed to catch: a record whose age exactly equalled the staleness window was treated as fresh, so it stayed authoritative for one tick past its own expiry. Freshness is now strictly inside the window. The parity test meant to stop the two guards' executor lists from drifting could never have failed. Its project fixture was a bare directory rather than a repository, so the non-git inert branch answered before the executor list was ever consulted. It asserted agreement it never actually measured. The fixture is now a real repository, like every sibling in the file. A test also asserted that Windsurf declares the worktree flag. It does not -- Windsurf resolves to no isolation by design, having no named concurrent dispatch to isolate. The test claimed a registry fact that was never true, and a comment in the resolver repeated it. Both corrected, and the test now proves what it should have all along: that the parser accepts any bare flag value, rather than one runtime's supposed value. The new guard was missing from the bundled-hook whitelist, which is the surface that decides what actually ships, and the per-plan gate had gained calls to the launcher without the preamble those calls require. The changeset carried parenthetical product descriptions the purity rule forbids. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#3045): backfill changeset pr number Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test(#3045): make the guard tests hold on Windows Two tests redirect HOME to control where the installer-persisted runtime default is read from. Node resolves the home directory from USERPROFILE on Windows and never consults HOME, so both silently read the real runner profile, found no recorded runtime, and asserted against a project the hook had not recognised. The production code was already correct in asking the platform rather than the variable; only the tests were wrong to assume one variable answers everywhere. The helpers now mirror the override onto both. The symlink spoofing test also created a directory symlink unconditionally, which needs elevated privileges on Windows. It survived on this runner, but it would fail on any host without them, so the creation is now attempted and the test skips explicitly when it cannot be done -- a bare return would have counted as a pass and hidden the gap. Skipping alone would have left the platform uncovered, so the behaviour it proves is now also driven in-process through an injected realpath, following the seam already used for the clock. That case no longer depends on privileges at all, and the end-to-end test keeps its original assertions wherever symlinks work. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-04 23:42:16 -04:00
Tom Boucher	ffd5370464	fix(#2903): use the command form that actually works in reader-facing docs (#3047) ... * fix(#2903): use the command form that actually works in reader-facing docs Docs told readers to type the colon form, which no runtime registers -- 18 of 19 runtimes use slash-hyphen and the 19th uses shell-var -- so anyone copying an example got an unrecognized command. Swept 178 occurrences across 53 files, locale mirrors included so they do not re-diverge from English. The colon form is a source-authoring token, not a user-facing one: install-time converters key on it to produce the hyphen form runtimes actually register. So the sweep is scoped, and three things are deliberately left alone: - ADRs, which are a historical record; editing their prose falsifies what was written at the time. - The legacy release-notes archive, pending a maintainer decision on whether it follows the same historical carve-out. Excluding it keeps a later reversal additive rather than a revert. - Source artifacts under commands, workflows and agents, where the colon form is load-bearing. Rewriting those would break the installed-skill guarantee across every runtime -- the single largest hazard here. The plugin namespace form is a real, separate token and survives untouched. Adds a lint enforcing exactly that boundary, since the correct form genuinely differs by directory and nothing previously caught the drift. Also fixes a hardcoded colon form in the capability-matrix generator. The sweep alone would have left the generated matrix disagreeing with the template that produces it, so the fix is at the source and the output regenerated. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#2903): stop the sweep misquoting source frontmatter Adversarial review caught three lines where the sweep rewrote a citation of the literal YAML name: key from a source command file. That key genuinely is the colon form -- this change's own carve-out logic says source-authoring tokens keep it -- so the docs ended up misquoting the real files. One of the three is an acceptance-checklist assertion, which the sweep turned into a false statement. Restored the three citations to match their sources verbatim, surgically: where a line carried both a name: citation and a real reader-facing slash command, only the citation reverted and the command stayed corrected. The guard needed the same distinction, or it would have flagged the restoration and reddened the build: a gsd:<cmd> token preceded by name: is a citation of a source token and is now permitted. The exemption is deliberately narrow -- a bare gsd:<cmd> anywhere else still fails -- with a test pinning that narrowness. Also makes the detection case-insensitive. Review found /GSD:next slipped through silently; no such casing exists in the tree today, so this closes a latent gap rather than fixing a live one. Swept the whole tree for further corrupted citations: none beyond the three. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#2903): retire the stale-next invariant and sweep next like every other command Maintainer decision on a genuine conflict between two contracts. Invariant #3054 banned the literal /gsd-next from user-facing docs because it named a retired workflow-advance command. But commands/gsd/next.md is a live command -- the state-aware smart-entry launcher -- and this issue requires docs to use the hyphen form every runtime actually registers. Both could not hold for this one command, so docs had been sidestepping the ban by keeping the colon form, which is exactly the defect this issue exists to remove. FEATURES.md already recorded the reassignment: the hyphen form "is not the retired workflow-advance command; it is reserved for the state-aware smart-entry launcher. Workflow advancement remains under /gsd-progress --next." With that reassignment the invariant's premise is obsolete and the guard now contradicts the documented command form, so it is retired with a comment recording why rather than deleted silently. next is now swept like every other command, and the earlier exemption added to the new guard is removed so nothing is special-cased. Four citations of the literal name: frontmatter key stay in colon form, because the source file really does carry name: gsd:next and a doc quoting it must reproduce it verbatim. Two of those lines were reworded to say which side is the frontmatter key and which is the slash command, since they previously conflated the two. Verified the retired scan would now genuinely fail against this tree -- the conflict was real and resolved, not dodged. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#2903): backfill changeset pr number Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-04 13:23:44 -04:00
𝚌𝚕𝚎𝚣𝚌𝚘𝚍𝚒𝚗𝚐	88f6d9bd1b	fix(#2644): deduplicate Cursor slash menu (#2812) ... * fix(#2644): deduplicate Cursor slash menu * fix: preserve installer executable mode * chore: add changeset for PR #2812 * test(#2644): acknowledge Cursor emission changes * test(#2644): drop spent emitted drift acknowledgments * fix(#2644): remove retired Cursor command converter --------- Co-authored-by: clezcoding <clezcoding@users.noreply.github.com> Co-authored-by: Tom Boucher <trekkie@nomorestars.com>	2026-08-03 12:05:45 -04:00