1094 Commits

Author	SHA1	Message	Date
Jakub Zych	792139b5ed	chore: sweep Kimi mentions from comments and notes	2026-10-06 20:19:17 +02:00
Jakub Zych	6cfa0c55d2	refactor: drop 12 runtimes, keep Claude, Codex, OpenCode, Cursor, ZCode, Antigravity ... Removes kilo, kimi, kimi-code, copilot, windsurf, augment, trae, qwen, hermes, cline, codebuddy and pi end to end: capability descriptors, installer branches and converters (bin/install.js 14.9k -> 11.2k lines), TypeScript converters, hook surfaces and runtime homes, review lanes qwen/kimi-code, the two pi migrations, Kimi payload normalization in the hook guards, dead hostBehaviors vocabulary, launcher home probes, fixtures, runtime-specific tests and the prose that presented them as supported. Installer output for the six kept runtimes is byte-identical to before the prune. The Kimi tool-vocabulary tests in workflow-guard, read-guard and read-injection-scanner are left in place pending a decision.	2026-10-06 20:02:40 +02:00
Jakub Zych	12ee75a509	chore: point MSD at git.golem15.com/golem15/msd-core ... The fork lives on the golem15 Gitea forge, not GitHub. Package identity now parses either host and derives in-place raw URLs for Gitea; the identity-drift lint accepts the new host; README drops GitHub-only badges and the npm quickstart in favour of the checkout installer.	2026-10-06 10:41:56 +02:00
Jakub Zych	a9a7a328e6	refactor: hard-fork GSD -> MSD (Make Software Done) ... Mechanical rename produced by scripts/msd-rename.cjs: gsd/Gsd/GSD -> msd/Msd/MSD across contents and paths, upstream package/repo coordinates -> @golem15/msd-core and golem15com/msd-core. Deep links into upstream history, sibling upstream packages, the GSD-2 import feature, CHANGELOG.md and .changeset/ are kept as-is. Hand edits on top: MSD block-letter banner and logos, LICENSE copyright line, package/plugin identity, regenerated lockfile, install-tree fixtures, derived registries and benchmark baseline; migration checksum baseline re-locked (MSD keeps its own install state, so no install had applied the old sums); sort-order and regex-escaped expectations in tests adjusted.	2026-10-06 01:47:40 +02:00
0xdhx	238bee7b03	fix(#4881): trust worktree.baseRef:"head" in harness mode; a WorktreeCreate hook withholds it and only a measured fork base restores it (#4921) ... * fix(#4881): trust worktree.baseRef:"head" in harness mode and keep the #4868 observation as what restores it under a WorktreeCreate hook The pre-dispatch base check still derived its harness-mode verdict from the retired #48 premise that the harness never reads worktree.baseRef: with "head" set and HEAD diverged from origin/HEAD it degraded every wave with baseref-head-ignored-by-harness. #4868 inserted an observation of a clean prior harness worktree at HEAD ahead of that comparison, but an execute-phase run never has one at the moment it checks — the base-check runs before any dispatch, a degraded wave creates no worktrees, and a wave that did run in worktrees has them removed and HEAD moved before the next check — so the common case was unchanged (#4881 repro states 1 and 4). Re-scope of the closed #4752 onto current next, with #4868 kept: - branch a trusts "head" in both isolation modes, the way the harness is measured to behave (#4588: three settings layers, three OSes), and the spawn-time exit-42 guard stays the observation-based backstop; - a Claude Code WorktreeCreate hook in any settings file the check reads, or a file that does not parse, withholds that trust — the hook creates the worktree without applying the setting — and the inferred comparison runs, degrading with baseref-head-bypassed-by-hook; - the #4868 observation (b2) now sits behind branch a: it is reached only when "head" was not trusted outright, and on a hook host it is what restores the trust — a hook that forks from HEAD leaves exactly that evidence, one that forks elsewhere never does. Its per-HEAD cache is unchanged. It is skipped under an explicit --observed-fork-base, which outranks an inference from a prior worktree; - --observed-fork-base <sha> threads a measured fork base through the evaluation (strict full-hex, TypeError otherwise). Tests: the #4868 rows are unchanged and still reachable (they run with the setting unset); the #4752 rows re-land, with the one exit-128 row updated to the degrade #4734 pinned since; a new #4881 block pins the start-of-run state (baseref-head, git never consulted), the hook + observation composition in both directions, and that an explicit observation skips the probe. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RPXQQPzHGintWtbhBoQCRS * docs(#4881): rewrite the eight surfaces that still state the harness ignores worktree.baseRef:"head" Every prose surface #4868 left untouched still asserted the retired #48 premise as verified fact, starting with the step file the orchestrator reads. Each now describes the measured behaviour, the WorktreeCreate-hook exception, the --observed-fork-base input, and the #4868 observation as what lifts the hook degrade; docs/CLI-TOOLS.md gains the fork-from-head-observed reason row #4868 did not document. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RPXQQPzHGintWtbhBoQCRS * chore(#4881): set changeset fragment pr to 4921 * fix(#4881): withhold the #4868 observation under the hook interlock The hook interlock this PR added withheld the worktree.baseRef:"head" trust but still let b2's prior-worktree observation restore it, and that observation cannot be attributed to the hook. The evidence is a clean agent worktree sitting at the orchestrator HEAD; nothing on disk records which creator left it there, so one the plain harness created BEFORE a WorktreeCreate hook was configured — with HEAD unmoved since — reads as evidence for the hook. It was the one fail-open branch in a mechanism documented as fail-closed. Keying the observation cache by hook configuration does not close it. observeHarnessForkFromHead has two legs: a HEAD-keyed cache and a live probe over .claude/worktrees/agent-*. A hook-keyed cache simply misses, and the miss falls through to the probe, which re-finds the same stale worktree and re-confirms. The probe takes no hook input at all. So the observation is not consulted under the interlock rather than re-keyed: on a hook host the only admissible positive signal is an explicit --observed-fork-base measurement of the dispatch in hand, and absent one the inferred comparison runs and a mismatch degrades with baseref-head-bypassed-by-hook, leaving the spawn-time exit-42 guard as the backstop. Scoped to the case branch a. declined to trust: "head" set AND a hook (or an unparseable layer) in the harness's path. With no "head" setting b2 is #4868's own arm and is unchanged, hook or not — re-scoping that trust is a separate question this PR does not open, and a test pins the boundary. Cost, stated: a hook host with "head" set, on a branch diverged from origin/HEAD and passing no observation, now runs sequentially. It still runs parallel when HEAD matches origin/HEAD. No workflow threads an observation today. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NuT9wTyrMaxjAPevqeH4YZ * refactor(#4881): drop the now-unused forkRef message-builder parameter buildMsgBaserefHeadIgnored stopped reading forkRef when the message was made mode-neutral, and the parameter was retained with `void forkRef;` for symmetry with its two sibling builders, which do read it. Symmetry is not reason enough to keep a dead parameter on a module-private function with one caller, so drop it (#4921 review). Behaviour is unchanged; the message text is pinned by an existing full-string assertion, which is what covers the only real risk here — transposing the two remaining arguments at the call site. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NuT9wTyrMaxjAPevqeH4YZ * test(#4881): exercise both FULL_SHA_RE alternatives at their boundaries The invalid-observation list pinned 39 and 41 hex around the 40-hex SHA-1 arm but left the 64-hex SHA-256 arm's own +/-1 boundary unexercised, which the repo's boundary-coverage convention asks for (#4921 review). Adds 63, 65 and a 64-length non-hex string. The regex already rejected all three -- this is coverage of correct behaviour, not a fix -- so it carries no negative control against a pre-fix base. That it is not vacuous was shown instead by widening the arm to {63,65}, under which the row fails by name. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NuT9wTyrMaxjAPevqeH4YZ * docs(#4881): finish the surface sweep the hook interlock owes Self-found by this round's own pre-push adversarial review, over six passes. Eight sites, four classes. FIVE were prose still asserting the stance the interlock overturned, which reads as live canon to anyone arriving cold: docs/CLI-TOOLS.md, docs/CONFIGURATION.md and gsd-core/references/planning-config.md each still said the hook degrade is lifted "unless/until a clean prior harness worktree is observed"; observeHarnessForkFromHead's own header still said a qualifying worktree "can only exist if the harness forked from HEAD"; and a test name still called the observation "required on a hook host" when it is now inadmissible there. My own sweep had grepped for "restores"/"lifts" and missed every one -- the ordinary failure of a grep, which returns what you thought to search for. The SIXTH is the same class one step worse: gsd-core/workflows/execute-plan.md still said flatly that Claude Code's isolation="worktree" "forks from origin/HEAD, not live local HEAD" -- in a paragraph THIS PR already edits, a few sentences after the clause it corrected. A tombstone makes only its own line clean; adjoining text asserting the dead stance is the other half of the same defect. Now qualified on the setting, with the hook exception named. The SEVENTH is a proof-strength overstatement that predates this PR, with a driven counterexample: a worktree created from an older base and since `git checkout --detach`ed onto HEAD is clean, sits at HEAD, and satisfies the probe identically, so "can only exist" was false. The worktree's own reflog does retain that original checkout -- the information is not lost, the probe simply does not consult it. The EIGHTH is that the header described only one of the function's two legs. A cache hit returns the prior conclusion without reading any worktree, so "the probe reads a worktree's present state" was true of the live probe and false of the cache. The header now separates them, and names the cache's blindness as a third reason the observation is inadmissible under a hook. Gaps seven and eight are inherited from #4868 and accepted there for the no-hook case. Nothing about the mechanism changes here; only what the header claims for it. No behavioural change -- comments, prose, and one test's registered name. Emitted-Drift-Ack-Growth: execute-plan.md — the Pattern A paragraph gained a qualifying clause: it stated flatly that Claude Code forks from origin/HEAD, which is the premise this PR retires, a few sentences after the clause the PR had already corrected. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NuT9wTyrMaxjAPevqeH4YZ --------- Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com> Co-authored-by: Tom Boucher <trekkie@nomorestars.com>	2026-09-23 20:15:28 -04:00
Tom Boucher	d7b5b2c2b6	fix(#4906): migrate the ROADMAP.md Plans: line onto the PlanningDoc seam — Phase 2 (#4933) ... * feat(#4906): migrate the ROADMAP.md **Plans:** line onto the PlanningDoc seam Phase 2 of epic #4906. Migrates the two writers of ROADMAP.md's Plans field onto the seam ADR-4910 locks, and deletes both bespoke regexes per Decision 2 — a correct copy of a rule the seam now owns is the same divergence risk as an incorrect one. src/phase.cts's mutateMilestonePhase carried planCountBodyPattern, one capture group, replace-to-end-of-line: this is #4852, still live before this change. src/roadmap.cts's cmdRoadmapUpdatePlanProgress carried the correct three-arm sibling (the #2853/#3584 correction) that phase.cts never adopted. Both now call findField/setFieldValue/serialize against a PlanningDoc parsed from the same milestone/phase-confined substring their existing withPhaseSection / replaceInCurrentMilestone wrappers already compute — those confinement wrappers are unchanged, only the field-write mechanism inside them moved. Verified end-to-end through the real commands against real fixtures, not against an isolated reimplementation of the classification logic: cmdRoadmapUpdatePlanProgress and cmdPhaseComplete both preserve a trailing human annotation across a real count rewrite, and both leave a bracketed human annotation (the #3584 Finding A discriminator) untouched. Found and fixed inline, in the already-merged src/planning-document.cts, rather than deferred: BOLD_FIELD_RE recognized only **Label:** (colon inside the closing bold). gsd-core/templates/roadmap.md ships every field, Plans included, as **Label**: (colon outside) -- migrating roadmap.cts onto the seam as it stood would have silently regressed real generated ROADMAP.md files back to the bug this migration exists to remove. Widened to recognize both spellings; deliberately NOT widened to a bare unbolded Label: form, which would register ordinary prose as a spurious field. roadmap.cts's writer also recognizes a bare singular/plural count (1 plan / 3 plans, no fraction) as an existing count token to overwrite, not template- placeholder or freeform prose -- the template's own single-plan-phase shape and #3584 Finding B's fix. Preserved exactly; this shape is easy to drop by only porting the more common fraction form. Two of Phase 2's three originally-cited defects turned out to be already fixed on next, independent of this epic, and are struck via a dated ADR amendment rather than silently narrowed: #4862 (stateReplaceField's own anchoring hardening already preserves sibling fields) and #4499 (spliceFrontmatter's per-key preservation already keeps block sequences byte-identical). Both reproduced against the built module before being struck, not assumed. STATE.md's field-write engine is re-scoped out of this phase entirely -- not because of its get_impact rating alone (measured the same way, the two sites THIS phase keeps are also CRITICAL, and an earlier draft of the amendment claimed otherwise without checking; corrected) but because updateCore is a multi-field transaction with frontmatter sync and preservation reconciliation that does not map onto PlanningDoc's node model, where migrating it would mean designing that model, not calling an existing seam function. Refs #4852 Refs #4906 Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> * fix(#4906): preserve a trailing annotation with no em-dash separator Test authoring surfaced a real regression against an existing #3584 fixture: `**Plans**: 0/1 plans executed (11-16 are gap closure from VERIFICATION)` -- a parenthetical annotation glued on with a bare space, no em-dash -- was left completely untouched by the migrated code instead of being rewritten with the count updated and the parenthetical preserved. Root cause: planning-document.cts's parseBoldFieldLine splits a field's value from its trailing annotation only on the literal " -- " separator. An em-dash-separated annotation already lives outside `value` in `trailingSpan`, untouched by setFieldValue regardless -- that path was never broken. A parenthetical with no em-dash has nowhere to go but inside `value`, and the migrated classification required the WHOLE value to match a count-token shape exactly, so this case fell into "leave untouched." Fixed in the migrated call sites, not in the seam: prefix-match the count token against the field's current value, then re-glue whatever textual suffix follows WITHIN that value onto the new count text before writing. Correct for both shapes with no special-casing -- the em-dash case's suffix-within-value is empty by construction (the annotation already lives outside value), the parenthetical case's suffix is exactly the glued content, preserved verbatim. Deliberately not fixed by widening planning-document.cts's separator grammar to also recognize a bare-space-then-parenthesis: that seam is already-merged and already-tested, and guessing at an open-ended set of annotation shapes at the seam level is exactly what isTemplatePlaceholder already avoids by staying caller-side. "What counts as a Plans-field count token" is domain knowledge about this one field. phase.cts's writePlansField had no arm-2/arm-3 classification before this migration -- its original regex unconditionally overwrote whatever value was present. That unconditional-overwrite behavior is preserved exactly for values with no recognizable count-token prefix; only the recognized-count case gained suffix preservation, matching what phase.cts actually did before. Verified end-to-end via the real cmdRoadmapUpdatePlanProgress and cmdPhaseComplete commands against real fixtures for both the parenthetical and em-dash shapes at both sites. Refs #4906 Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> * test(#4906): cover the migrated Plans-line writers at both sites 15 cases across tests/phase.test.cjs and tests/roadmap.test.cjs, one per row of the phase test matrix, extending the existing describe blocks and helper functions those files already use for these two commands rather than building parallel fixtures. Covers: trailing-prose preservation on a real count rewrite at both sites (the #4852 regression, and the #2853/#3584 non-regression); zero-trailing- content boundary; the bracketed-template-placeholder vs bracketed-human- annotation discriminator (#3584 Finding A); a missing Plans field not crashing either command; an unrelated unreadable sibling node in the same confined section surfacing rather than corrupting the field; confinement holding across sibling phases and milestones; a round trip through the command's own read path; CRLF safety; and a parity assertion that both sites now produce identical Plans-line text for identical inputs, proving one shared mechanism rather than source-grepping for the deleted regex literals. Authoring caught a real regression before it could land silently: an existing #3584 fixture using a parenthetical annotation with no em-dash separator failed against the first version of the migration. Reported rather than edited to match the broken behavior -- see the paired fix commit. That existing test needed no changes once the fix landed; its assertion was verified independently against the real CLI before this commit. Refs #4906 Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> * fix(#4906): give phase.cts's Plans writer the same arm-2/arm-3 classification as roadmap.cts Isolated adversarial review (mandatory orthogonal review pass) executed writePlansField against `[Deferred pending re-scope]` and the fresh-template placeholder wording and found the first version of this migration kept phase.cts's OLD unconditional-overwrite behavior for the no-count-prefix case instead of adopting the same isTemplatePlaceholder / arm-3-untouched classification roadmap.cts's sibling site already uses. A bracketed human annotation was being silently rewritten to a new count -- a real content-destroying regression against this phase's own design-doc behavior table, not an accepted trade-off, and exactly the kind of divergence between the two sites Decision 2's parity requirement exists to eliminate. writePlansField now runs the same template-placeholder check and "no count prefix and not a placeholder => leave untouched" branch before ever calling setFieldValue. Added two site-1 tests (rows 4 and 6 of the phase test matrix) mirroring the existing site-2 coverage for this exact discriminator. Refs #4906 Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> * fix(#4906): restore plain (non-bold) Plans: line support and fix a test helper mismatch gsd-test (mandatory verification, run before every push) surfaced two real defects the migration's manual CLI checks had not caught: 1. #1163 regression: a hand-edited/pre-template ROADMAP.md can carry a PLAIN (non-bold) `Plans:` line rather than the canonical `**Plans**:`/ `**Plans:**` bold field. The old deleted regexes tolerated this shape; the seam's BOLD_FIELD_RE is deliberately bold-only (widening it would register ordinary prose like "Note: see below" as a spurious field seam-wide), so the migrated writers silently no-op'd on it instead of updating the count -- a real, previously-tested behavior lost. Fixed with a caller-side fallback in both src/roadmap.cts (where the failing #1163 test lives) and src/phase.cts (added for parity, per Decision 2 -- the two sites should not diverge on which legacy shapes they tolerate): when findField finds no boldField Plans node, look for a plain `Plans:` line directly and apply the same arm-1/2/3 classification against it. This is domain knowledge about one field's legacy tolerated shape, the same class of thing isTemplatePlaceholder already keeps caller-side rather than seam grammar. 2. tests/roadmap.test.cjs's new rows 4/5 (site 2) seeded the colon-outside spelling (`**Plans**: ...`) but their plansLineIn() helper only matched colon-inside (`**Plans:**`), so both assertions compared against `undefined` regardless of whether the write logic was correct -- a test-authoring bug, not a source defect. Fixed the helper to recognize both BOLD_FIELD_RE spellings, matching what the production code actually supports. Verified via a real gsd-test run before this fix (outcome: failed, 5 failures, all in tests/roadmap.test.cjs) and will be re-verified via a real gsd-test run on this commit before push, per this repo's non-rationalization rule: a red gate is fixed, never explained away. Refs #4906 Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> * chore(#4906): backfill the Phase 2 changeset fragment's PR number pr:0 -> pr:4933, now that gh api POST /pulls has returned the real number. Refs #4906 Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>	2026-09-23 19:25:39 -04:00
Michel Moreira	1e3e1f7cd8	enhance(#4570): allow disabling planner stall detection (#4585) ... * enhance(#4570): allow disabling planner stall detection * docs(#4570): add changelog fragment Emitted-Drift-Ack-Growth: plan-phase.md — the explicit opt-out gate covers all five planner and checker spawn classes Emitted-Drift-Ack-Growth: settings-advanced.md — the toggle prompt and bounded-recovery warning expose the new setting * chore(#4570): refresh compact-content baseline * fix(#4570): preserve default-on watchdog fallback * docs(#4570): qualify the chunked-mode orchestrator rules with the toggle The two chunked-planning-mode stall-watch imperatives read as unconditional, with the opt-out stated only in a following bullet. State the PLANNER_STALL_DETECTION_ENABLED condition inline, matching the three sites already qualified in plan-phase.md. * chore(#4570): refresh compact-content baselines against rebased next * fix(#4570): sync planner stall launcher Keep the stall-detection helper aligned with the canonical runtime launcher. * chore(#4570): refresh compact baseline after rebase --------- Co-authored-by: Tom Boucher <trekkie@nomorestars.com>	2026-09-23 18:46:23 -04:00
Michel Moreira	76ef60ba25	enhance(#4836): prefer the graphify CLI for planner and researcher graph queries (#4874) ... * enhance(#4836): prefer the graphify CLI for planner and researcher graph queries The planner gets one knowledge-graph query per phase and the researcher two or three, and that single shot decides which modules the plan treats as related — and therefore how tasks are ordered into waves. It was spent on the built-in reader, which seeds by case-insensitive substring match over a node's label and description and then expands a hardcoded two hops. The phase "User Authentication" seeds on `author`, `authoring` and `unauthorized` with the same weight as `authenticate`, and when the inflated payload exceeds `--budget` the trimmer drops edges by confidence tier — so the highest-confidence tier can be discarded to fit a payload that bad seeding inflated in the first place. The graphify CLI is already a hard dependency of /gsd-graphify build, and it ranks seeds (IDF weighting, trigram fuzzy matching) and applies context filters before traversal. Both prompts now prefer it and fall back to the built-in reader, branching on `command -v graphify` — the same degradation shape the repo already uses for Context7 to ctx7. Binary presence is a self-satisfying gate: a graph can only exist if the binary built it, so the fallback covers edge cases (a CI checkout with a committed graph, a binary since removed), not the common path. No new config key and no new tool grant — both agents already have Bash. The planner additionally runs `graphify affected`. The reference states its own goal as "which subsystems may be affected by changes in this phase", which is literally reverse traversal by relation; the built-in reader only approximates it with undirected two-hop expansion and has no equivalent verb, so `affected` is skipped on the fallback path. `graphify status` now reports `graph_path`, the resolved absolute graph location, on both the present and the missing branch. The CLI takes the graph location as `--graph`, and the prompts must not re-derive `.planning/graphs/graph.json` for it: that would point the CLI at a non-existent local mirror in exactly the umbrella multi-repo setup `graphify.graph_path` (#1825) exists to serve. For the same reason the presence gate in both prompts is now the `status` call itself rather than a bare `ls` of the default location, which was already blind to the override. Known limit, stated in both prompts rather than implied: the two paths return different shapes. `graphify query` emits prose and has no `--json` flag; the built-in emits JSON with per-edge confidence tiers and budget_met/budget_estimate. `--budget` also counts rendered output on one and estimated payload bytes on the other (#2738) — same flag name, different unit. Both are read by a model and nothing machine-parses the injected block. With graphify absent from PATH the injected context is byte-identical to before. Closes #4836 Emitted-Drift-Ack-Growth: gsd-phase-researcher.md — the CLI-first branch, the reason it is preferred, and the output-shape warning are the deliverable; a pointer to a part would not be read at the decision point. Emitted-Drift-Ack-Growth: gsd-planner.md — one sentence in the load_graph_context step pointer, so it stops naming the default graph path the reference no longer assumes. * docs(#4836): record the CLI-first graph query in the planner and researcher entries * chore(#4836): add changeset fragment * enhance(#4836): name the full domain word in the planner's query-term examples The reference's own example — phase "User Authentication" → term "auth" — is the exact collision the CLI-first path exists to avoid, and it stays a collision whenever the fallback path runs, since that path matches the term as a substring of label and description. * fix(#4836): surface graph_path on the unparseable-graph status branch graphifyStatus() returned graph_path on the exists:true and exists:false outcomes but not on the third, error, outcome (graph.json present but unparseable). The planner/researcher prompts gate CLI-first dispatch on exists, not on this outcome, so a corrupt graph file made them fall through to the CLI-first branch with the literal <graph> placeholder and no real path to substitute. * docs(#4836): note graph_path's trust boundary at the --graph interpolation graph_path is reflected verbatim into a double-quoted --graph argument the agent executes via Bash. It comes from graphify.graph_path, a config surface already trusted elsewhere, so this isn't a new trust boundary -- but it is a new injection site (no --graph flag existed on this call before). One-line caution for anyone hardening this later. --------- Co-authored-by: Tom Boucher <trekkie@nomorestars.com>	2026-09-22 19:59:45 -04:00
Tom Boucher	585cab41cc	fix(#4770): lift the Codex sandbox_mode holds — documented enforcement suffices (#4920) ... * fix(#4770): lift the Codex sandbox_mode holds — documented enforcement suffices * docs(#4770): backfill the changeset PR number --------- Co-authored-by: sim <sim@local>	2026-09-21 15:30:52 -04:00
Tom Boucher	6a4984cf69	fix(#4763): surface the displaced session record and pass --phase from the executor decision loop (#4919) ... * test(#4763): failing-first — replaced-record payload and executor --phase pins * fix(#4763): surface the displaced session record and pass --phase from the executor decision loop state record-session keeps its last-writer-wins write (the recorded single-slot handoff design) but no longer displaces silently: when a non-empty Stopped At or authored Resume File record is replaced, the payload carries the full prior text under replacedRecord. Same-value rewrites, the insert path, and the #944 template-default DWIM are not displacements and report nothing. The executor decision loop now passes --phase "${PHASE}" to state.add-decision, matching execute-plan.md, so decisions stop inheriting whichever phase the global pointer names (#4763 case 2). advance-plan is unchanged (#3311 by-design). Emitted-Drift-Ack-Growth: gsd-executor.md — the decision loop gained its --phase guard and a comment naming why (#4763) * docs(#4763): add the changeset fragment * docs(#4763): backfill the changeset PR number --------- Co-authored-by: sim <sim@local>	2026-09-21 12:42:11 -04:00
Tom Boucher	55fba5f7ce	feat(#4917): add the PlanningDoc parse → mutate → serialize seam — Phase 1 of #4906 (#4918) ... * feat(#4917): add the PlanningDoc parse -> mutate -> serialize seam Phase 1 of epic #4906, implementing ADR-4910 and its 2026-09-21 amendment. Net-new leaf module; NO call site is migrated, so nothing in the twelve absorbed issues changes behavior yet. src/planning-document.cts composes the seams that already exist rather than reimplementing them: markdown-sectionizer for structure (fences and code spans come from stripFencedCode / scanInlineCodeSpans, never a second scanner), markdown-table for tables, frontmatter for frontmatter, write-set for Result<T>. What is structural rather than conventional: - A field node carries labelSpan, valueSpan and trailingSpan separately, and the only write entry point takes a node id and writes into valueSpan. trailingSpan is readable and has no exported writer, so the #2853/#3584/#4852 rule ("the verb owns the count token ONLY") stops depending on an author remembering a third capture group. - Mutation is node-addressed. A handle is minted by the parser, so a caller cannot name a node the parser did not find. No path strings — a path is a grammar, and a grammar needs a parser. - serialize splices staged spans into the ORIGINAL buffer. A no-edit serialize is byte-identical, which is what eliminates the #4499 defect without a targeted fix, and which also makes an already-escaped table cell impossible to double-escape on a round trip. - A node that fails to parse carries its own error and span; siblings stay readable. - Per the ADR amendment, serialize REFUSES whenever any node carries a parse error, even with zero staged edits, naming the offender. One defect was found while building and fixed in place rather than deferred: PLANNING_ARTIFACTS first derived from isCanonicalPlanningFile unfiltered, so config.json, state.json, milestone.lock and skill-manifest.json were accepted and returned {ok:true, nodes:[]} — "this document records nothing" when the truth was "I have no grammar for this file". That is the exact empty-vs-error confusion this epic exists to remove (#4899, #4900), reproduced inside the seam built to prevent it. The registry now filters to .md while still DERIVING from isCanonicalPlanningFile, because hand-writing a second list is the divergence this epic is about, and parsePlanningDoc refuses a non-markdown kind at the document level per ADR-4910 section 5. New bin/lib module bookkeeping: .gitignore, eslint.config.mjs ignore (ADR-457 — lint the .cts, not the emitted .cjs), docs/INVENTORY.md row, regenerated docs/INVENTORY-MANIFEST.json, and the CONTEXT.md glossary entry. Refs #4906 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test(#4917): cover the PlanningDoc seam across 28 input classes 30 cases in 14 describe blocks, one per row of the phase test matrix. The two load-bearing tests are the fast-check properties (seed 20260921, numRuns 200): - a single node mutation leaves every byte outside that node's valueSpan identical to the source - serialize with zero staged edits is the identity function Both are DOCUMENT-SHAPED per CONTRIBUTING.md fixture provenance (#2371): the generator assembles arbitrary frontmatter, heading, label and value text with join(), and never calls serialize or any other function from the module under test to build a fixture. Seeding the generator from the module's own writer would make the document shape a constant, and the property could then never explore a document the writer would not itself emit. Verified the byte-range assertion is not vacuous with a control run: it passes against the real writer and FAILS against a simulated #4852 writer (one capture group, replace-to-end-of-line), which visibly drops the trailing annotation. Boundary coverage is zero / one / two staged edits. Negative space carries its own rows — bold emphasis in prose, a field-shaped line inside a fenced block, the same inside an inline code span, and a horizontal rule mid-body all correctly mint no node. Row 24 is the Generative-Fix-Divergence parity assertion: PLANNING_ARTIFACTS must not diverge from isCanonicalPlanningFile. Row 28 covers the non-markdown canonical file found during the build. Assertions are structural throughout — a typed Result / NodeRead / SerializeOutcome shape, or a byte range computed from the node's own Span. Full-string equality appears only where the contract IS byte equality. Refs #4906 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#4917): apply review findings — refuse unrepresentable values, compose the layers the seam claimed Three review passes ran against this branch: /security-review, an isolated adversarial pass, and a standards+spec pass. Five findings, all fixed here. Every one is the epic's own failure class reproduced inside the seam built to end it, which is the thing worth noticing. 1. setFieldValue accepted a value containing a newline. It survived serialization and reparsed as a REAL sibling field — one write to Plans forged a second Owner into a document that already had one. Content became structure, which defeats ADR-4910 Decision 2's "cannot reach past its own token by construction": the token boundary is a LINE boundary. 2. setFieldValue accepted a value containing the trailing separator and SILENTLY TRUNCATED it. Staged "sneaky - annotation", read back "sneaky", with the remainder reclassified as trailing prose. No error, nothing unreadable, both resulting nodes parsing perfectly. Worse than (1) because it loses the caller's own value rather than adding something visible. Both are fixed by ONE general check, deliberately not a blacklist: setFieldValue rebuilds the candidate line, re-parses it through the same field grammar, and refuses unless the value reads back identical. Blacklisting the separator would close this instance and leave the class open for whatever separator the grammar grows next. The round-trip check is ADR-4910 Decision 4 stated executably. 3. The module reimplemented two layers it claims to compose. Checklist detection hand-rolled a checkbox regex that markdown-sectionizer's iterateBullets already owns. Frontmatter span detection re-derived fence handling because frontmatter.cts's frontmatterRegion was module-private — so ADR-4910 section 1's stated layering was UNREACHABLE as written, and the first implementation routed around it silently instead of surfacing the gap. frontmatterRegion is now exported (additive only; ADR-2143 section 2's extend-never-mutate lock is inherited) and both layers are consumed. 4. The CONTEXT.md glossary entry asserted "the Frontmatter Module supplies frontmatter" while zero frontmatter.cts code was invoked. That was a false claim in the repo's vocabulary of record, written by me, and it is now true rather than edited away. 5. Adopting iterateBullets narrowed GFM coverage: it classifies only dash-prefixed task items as checkboxes, so "* [ ] x" and "+ [x] y" stopped becoming checklist nodes. Widening the sectionizer is forbidden by the inherited lock, so the task-list MARKER is interpreted in this module while bullet STRUCTURE still comes from the sectionizer. The sharpest finding was not a defect. The fast-check generator constrained values to [A-Za-z0-9 .,!?], so it could not emit an em-dash, newline, backtick, pipe or asterisk — precisely where (1) and (2) lived. The property was real, seeded and non-vacuous, and structurally blind to the module's actual bug class. The generator now spans the grammar's own metacharacters, and a new property asserts that every value setFieldValue ACCEPTS round-trips identically. Proven able to fail: against a scratch copy with the guard stripped it fails after 7 cases on newValue "\n". Recorded as a measured boundary, not fixed: a bare CR inside a field line leaves that field unrecognised. Measured — sibling fields still parse, no error node, and serialize stays byte-identical, so the worst case is an unreadable field and never a damaged document. Flagged for Phase 4's empty-vs-error census. Refs #4906 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#4917): backfill changeset pr number to 4918 Refs #4906 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-09-21 11:17:58 -04:00
Tom Boucher	eea9247c93	enhance(#4095): checkpoint:decision auto-selection is opt-in via auto_select (#4912) ... * enhance(#4095): checkpoint:decision auto-selection is opt-in via auto_select Auto-mode used to auto-select a checkpoint:decision's first <option> unconditionally, making a decision checkpoint's safety depend on option presentation order rather than an authored choice. Add an optional auto_select="<option-id>" attribute on the <task> tag: absent, auto-mode now escalates to a human exactly like gate="blocking-human" does; present, it names the option auto-mode selects; naming an id with no matching <option id> is a hard structural-validation error at plan-parse time rather than a silent fallback to the first option. gate="blocking-human" continues to win over everything, unchanged. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> * fix(#4095): anchor auto_select/id attribute regexes past hyphenated decoys An isolated adversarial review of the auto_select work found that both new attribute regexes used \b as their left anchor, which is a word boundary, not a "start of attribute name" boundary. A decoy attribute ending in the same word (e.g. data-id="...") sitting before the real id="..." on the same <option> tag matched first, silently corrupting the extracted option id. Anchor on (?:^|\s) instead so only the real attribute name can match. Adds a regression test reproducing the exact decoy-attribute shape, plus a Unicode option-id test from the same review pass. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> * test(#4095): register auto-select-attribute.test.cjs in the docs-guard lane lint-docs-guard-registration failed: the new test reads docs/reference/ plan-md.md but was not registered, so a future edit to that doc could silently desync from the test without the guard catching it on the PR that changed the doc. Registered alongside its direct precedents (precondition-element.test.cjs, reversibility-tagging.test.cjs), which read the same file for the same reason. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> * fix(#4095): fit the decision bullet under execute-phase.md's frozen byte ceiling The remote gsd-test run caught what local checks missed: execute-phase.md carries a frozen ADR-857 Phase-6 byte ceiling (93600) with only 36 bytes of headroom before this change, and the original checkpoint:decision wording pushed it to 93772 (over the ceiling). Cascaded into failures in phase6-capstone-conformance, execute-phase-completion-reconciliation, claude-orchestration, and the compact-content drift-report test. Also caught: tests/package-legitimacy-gate.test.cjs anchors a "decision is conditional, not unconditional" safety check on the literal phrase "first option" in the decision bullet — which #4095 deliberately removes, since there is no more unconditional first-option pick. The test was asserting an assumption this change intentionally makes obsolete; re-anchored on tokens that still identify the bullet ('decision', 'auto-spawn') without weakening what the test actually verifies (the bullet must still carry a blocking-human carve-out). Also fixed a word-order mismatch between my own new test's regex and the actual doc text it was asserting against (tests/auto-select-attribute.test.cjs). Regenerated the compact-content benchmark baseline (tests/fixtures/compact-content-benchmark-baseline.json) to match the new byte counts. Emitted-Drift-Ack-Growth: gsd-executor.md — +7 bytes (49138 -> 49145), from the auto_select carve-out added to the checkpoint:decision auto-mode bullet; already trimmed once to fit the 49152 hard cap. Emitted-Drift-Ack-Growth: execute-phase.md — +12 bytes (93564 -> 93576), from the same carve-out in the orchestrator's decision bullet; kept 24 bytes under the frozen 93600 ADR-857 ceiling after two rounds of trimming for clarity vs. margin. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> * chore(#4095): backfill changeset PR number Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>	2026-09-20 23:47:01 -04:00
Tom Boucher	ccfed63355	fix(#4823): the Current Plan reset is scoped to the Current Position section (#4898) ... * test(#4823): failing-first — the Current Plan reset must not rewrite prose outside Current Position * fix(#4823): the Current Plan reset is scoped to the Current Position section — the whole-body 'Plan' fallback matched hard-wrapped prose lines starting with plan: * chore(#4823): changeset fragment * chore(#4823): backfill changeset PR number (4898) --------- Co-authored-by: sim <sim@local>	2026-09-20 11:45:50 -04:00
Tom Boucher	bd79a97df0	fix(#4806): unparseable VERIFICATION.md frontmatter reports a distinct parse error, never status missing (#4896) ... * test(#4806): failing-first — unparseable VERIFICATION.md frontmatter is a parse error, not status missing / Field not found * fix(#4806): unparseable VERIFICATION.md frontmatter reports a distinct parse error — never 'missing' or 'Field not found' * test(#4806): census pin 66→67 — cmdFrontmatterGet's unparseable-frontmatter error is a new output({error}) call site * chore(#4806): backfill changeset PR number (4896) --------- Co-authored-by: sim <sim@local>	2026-09-20 10:26:53 -04:00
Tom Boucher	e2d879f681	fix(#4802): audit acknowledge refuses targets whose frontmatter fails to parse (#4895) ... * test(#4802): failing-first — acknowledge must refuse an unparseable-frontmatter target instead of splicing over it * fix(#4802): audit acknowledge refuses targets whose frontmatter fails to parse — never splices the marker-marked object over the file * chore(#4802): backfill changeset PR number (4895) --------- Co-authored-by: sim <sim@local>	2026-09-20 08:45:54 -04:00
Tom Boucher	a8394713d1	fix(#4801): init.manager resolves archived phase directories through findPhaseInternal (#4893) ... * test(#4801): failing-first — an archived phase directory must resolve and report complete in init.manager * fix(#4801): init.manager resolves archived phase directories through findPhaseInternal The private current-milestone-only scan (matchPhaseDirs over listMilestonePhaseDirs) counted archived phase directories as missing, so an archived phase with a passed verification reported no_directory/phase_complete:false. findPhaseInternal — already imported, already the shared primitive for five other init commands — searches the live directory first and falls back through the workstream-scoped archive (#2855); the private copy and its single-consumer entries list are retired. * chore(#4801): backfill changeset PR number (4893) --------- Co-authored-by: sim <sim@local>	2026-09-20 06:45:38 -04:00
Behruz Nassre Esfahani	2e14b4df17	fix(#4415): treat an absent worktree as removed, not as a branch mismatch (#4612) ... * fix(#4415): treat an absent worktree as removed, not as a branch mismatch Claude Code removes a subagent's worktree the moment the subagent finishes with a clean tree. A gsd-executor that committed everything — SUMMARY.md included, under `commit_docs: true` — is exactly that case, so by the time the orchestrator reaches wave cleanup the directory is routinely gone while the branch it left behind is intact and mergeable. `git -C <gone> rev-parse --abbrev-ref HEAD` fails, and nothing distinguished that filesystem failure from a real branch disagreement: both reached the same `if`, so the entry blocked `branch_mismatch`, NOTHING merged, and the branch was left dangling. When the directory instead vanished after the merge landed, `git worktree remove` failed "is not a working tree" and the entry blocked `worktree_remove_failed`, leaving the branch undeleted and the operator to run `git worktree prune` + `git branch -D` + `rm -rf` by hand every wave. Disambiguated at the point of failure rather than ahead of it. A SUCCESSFUL in-worktree read still decides identity exactly as before — a present worktree on the wrong branch blocks, unchanged — and only a FAILED read consults the filesystem. Two reads can fail, and they are not the same path: * The branch read fails with the directory absent. There is no checkout for identity to come from, so it falls back to `refs/heads/<branch>` read from repoRoot; a missing ref still blocks, so an absent worktree never becomes a silent pass. The SUMMARY rescue and the dirty check are then skipped. * The branch read succeeded and the later `status` read fails with the directory now absent — the harness removed it while the repoRoot-side base, deletion and scope checks ran. Identity was already established from the checkout and the rescue has already run; only the dirty decision is skipped. Without this, a mid-entry removal still blocked `worktree_dirty` with nothing merged: the same bug, one window later. Skipping those reads is not a claim that the worktree was clean. This code cannot tell who removed the directory, and a forced or manual `rm -rf` of a DIRTY worktree would already have destroyed an uncommitted SUMMARY before cleanup ran. The narrow thing that is true either way is that a missing source cannot be read. The two reads also fail differently: the default SUMMARY finder catches the unreadable directory and returns no files, while `git -C <gone> status` errors — and that error is what surfaced as `worktree_dirty`. A rescue that genuinely FAILS still blocks, since a copy that errored part-way can mean an uncommitted SUMMARY was really lost. Teardown prunes the stale .git/worktrees admin entry rather than removing a path that is not there, re-reading presence instead of reusing the branch-step answer since the harness can act in between. For an entry accepted as ABSENT it prunes ONLY and never issues `worktree remove --force`: that entry was merged without the rescue and dirty checks, so force-removing a checkout recreated at that path would delete contents that never passed either one — strictly worse than the bug being fixed. A genuine prune failure still reports `worktree_remove_failed`, and a blocked teardown still withholds the branch delete. `git worktree prune` is repository-wide maintenance, not an entry-scoped operation. The presence probe resolves `worktree_path` against repoRoot, the way git does. `normalizeCleanupManifestEntry` takes the path from the manifest verbatim, so it can be relative, and every git call passes it as `-C <path>` with `cwd: plan.repoRoot`; a bare `fs.existsSync` would have resolved it against the PROCESS working directory instead. Those differ whenever cleanup runs from elsewhere, reachable today through gsd-tools' `--cwd` override, and the mismatch reads both ways: a present checkout reported absent — skipping the dirty check that would have blocked it — or an absent one reported present. An earlier cut resolved presence UP FRONT, before the branch read. That broke 52 existing tests: every cleanup-wave test uses a fake path that does not exist on disk and injects no `existsSync`, so all of them re-routed down the absent branch. Disambiguating at the point of failure leaves those tests reading as they did. Three rows still needed their premise stated — each stubs a git failure against a worktree that is genuinely present — and now inject `existsSync: () => true`. No assertion in any of the three changed. Fourteen rows added. Every early row held presence CONSTANT and so could not reach the windows that matter, since the bug is caused by a directory that changes state WHILE cleanup runs: removal after the branch read, a present worktree whose status fails (which must still block), removal between the clean status read and teardown, a reappeared checkout at teardown, #2852 isolation of a blocked absent entry from the entries after it, and relative-path resolution. Verified: ran the issue's own reproduction verbatim against a build of this branch — `merged_removed`, merge commit present, branch deleted, no prunable entry in `git worktree list`. The same reproduction against a build at the merge-base returns blocked/branch_mismatch, no merge, branch present, `wt1 ... prunable`. Five of the first eight rows go red against the true merge-base file; the three that stay green are the safety-preservation rows. The rows added after each review round go red against the commit that round reviewed. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NRaNKCDUEacHudVDwvat8X * chore(#4415): add changeset Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NRaNKCDUEacHudVDwvat8X * fix(#4415): build the probe-path expectation with path.resolve, not path.join The row asserting that the presence probe resolves a relative `worktree_path` against repoRoot failed on windows-latest while the code under test was correct. On win32 `path.resolve` prepends the current drive to a drive-less absolute path (`/repo/main` -> `D:\repo\main`) and `path.join` does not, so a join-built expectation disagrees with correct behavior: expected: '\repo\main\.claude\worktrees\agent-a1' actual: 'D:\repo\main\.claude\worktrees\agent-a1' `path.resolve` is what the fix must use — it is how git resolves `-C <path>` against `cwd: plan.repoRoot` — so the expectation moves to resolve as well. Two `notEqual` rows keep that from being circular: the probe must receive neither the raw relative path nor a process-cwd resolution. Verified by mutation — dropping the repoRoot anchoring in `worktreeExists` turns the row red. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NRaNKCDUEacHudVDwvat8X * fix(#4415): confirm absence before skipping the rescue and dirty checks `fs.existsSync` answers false for a genuinely missing path AND for one it merely cannot traverse — EACCES on a parent directory, an unreachable mount. Verified: with a parent at mode 000, `existsSync` returns false while `statSync` throws EACCES. That distinction carries weight here, because "absent" is what lets an entry skip the SUMMARY rescue and the dirty check. An unreadable-but-present worktree read as absent, so cleanup merged over uncommitted work that the dirty check exists to refuse — and it contradicted this code's own comment that a present checkout whose git read fails stays blocked. Before this PR a failed git read blocked unconditionally, so treating unreadable as present is not a new safety rule; it is the one that was already there. The default probe becomes `statSync`, which reports WHY it failed. Only ENOENT is absence; anything else reads as present and blocks. An injected probe stays authoritative, so tests state presence directly with no hidden dependency on the real filesystem, and may throw to state that a path is unreadable. Two rows added: an unreadable worktree still blocks as branch_mismatch with no merge and no teardown, and a confirmed-ENOENT probe still takes the absent path. Verified by mutation — reverting the discrimination to the permissive `return false` turns the unreadable row RED while the ENOENT row stays green, which is what distinguishes discrimination from over-blocking. The mutation was confirmed to reach the compiled artifact the test loads. Also from this round: the row named for a checkout that "reappeared" never modeled reappearance (production probes presence once, at identification), so it is renamed to the unconditional contract it does prove; the comment crediting the notEqual rows with removing circularity is narrowed to what they actually establish; and the changeset now says only a confirmed absence takes the new path. Found by Codex full-PR review (round 3) before pushing. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01S4mJZpSNwoyVtRVUQoijfL * fix(#4415): source identity from git's registration, removal from the errno Maintainer review rejected the premise this fix rested on. It held that once the worktree directory is gone there is no checkout to read, so identity must fall back to `refs/heads/<branch>`. Git does not lose the binding — measured, after `rm -rf`: worktree /path/to/wt branch refs/heads/feat-x prunable gitdir file points to non-existent location The ref fallback weakened identity from "the checkout registered at this path is on this branch" to "a branch by this name exists", which let a foreign sibling branch merge. Identity now comes from `git worktree list --porcelain`, so the #3677 swap control keeps its teeth on the absent path; the new swap row is what would have caught this, and dropping the branch conjunct turns only that row red. Two defects in the first cut of the porcelain rework, both measured rather than reasoned about: `prunable` is not a removal test. With a parent directory at mode 000, git prints `prunable gitdir file points to non-existent location` for a checkout that is STILL THERE — it cannot traverse the parent, so it reports the gitdir file as missing. Treating prunable as "removed" would skip the rescue and dirty checks and merge over uncommitted work in an unreadable worktree, reintroducing the review's Major finding by another route. Each source now answers only what it can prove: porcelain for identity, `statSync`'s errno for removal. Only ENOENT is removal; EACCES/EIO blocks, as it did before this PR. `git worktree prune` is repository-wide. Measured: two removed worktrees plus ONE prune leaves neither registration behind. Reading the list per entry therefore let the first absent entry's teardown erase the identity evidence of every entry after it, merging one worktree per wave and blocking the rest as branch_mismatch — worse than the bug being fixed, since a wave of parallel executors is the normal case. The identity read is now a snapshot, captured lazily on the first entry that needs it and reused for the wave, which is both pre-prune and off the happy path. The `existsSync` probe and its dep locals are deleted; the filesystem is consulted only for the errno. The comment calling repository-wide prune "Harmless" was wrong under the new identity rule and says so now. Tests: identity and removal are stated on their own axes rather than through one present/absent boolean. Added the absent-path #3677 swap row, the two-absent-entry prune row, a bare `prunable` marker row, and a fail-safe row for an unreadable worktree list. Three mutations each kill exactly the intended rows, verified against the compiled artifact the tests load. One fixture that still stated presence through the removed `existsSync` seam was passing for the wrong reason and now states both axes. Verified: lint:ci exit 0; full suite 24/24 chunks, 37,164 tests, 0 failures; tests/worktree-safety.test.cjs 422/422. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01S4mJZpSNwoyVtRVUQoijfL * fix(#4415): re-confirm absence before teardown, and prove the porcelain claim against real git Maintainer review, Major. Presence was classified once, at identification, and everything between that point and teardown — the base, deletion and scope gates, and the merge itself — is a window in which a worktree can reappear. The defence was "prune only, and a live checkout would make `branch -D` fail visibly", which holds only while prune's own staleness check is not fooled by the same filesystem-visibility gap that produced the false absence one call earlier. If it is, prune clears the admin entry, `branch -D` then SUCCEEDS, and a live, unreviewed, un-rescued worktree loses its branch. That asymmetry is the argument for the fix: the bug this PR set out to repair only ever BLOCKED, while this path could DESTROY state. Absence is now re-confirmed with `confirmedGone()` immediately before teardown — no new subprocess, just the statSync already in hand — and a reappeared directory blocks as `worktree_remove_failed` instead of reaching prune or the branch delete. The review was also right that the gap was known and unverified: the existing row said so in its own comment ("it does NOT model the reappearance transition itself"). It is modelled now, by a stat that answers "gone" at identification and "present" at teardown. Mutation-verified: removing the re-confirmation turns ONLY the new row red while the old "prune, never force-remove" row stays green, which is exactly why that row could not have caught this. Minor, same review: the #4415 block was entirely mock-based, so the factual claim the identity mechanism rests on was asserted in comments and measured out of band but never proved executably. Two real-git rows now prove it — that git keeps the path -> branch binding after the checkout is deleted and marks the entry prunable, and that it ALSO reports prunable for an unreadable worktree that is still there, which is why removal is confirmed by errno rather than by prunable. The second row skips as root, where mode 000 does not deny traversal. Minor 2 (rescueSummaryArtifacts resolving worktree_path against process.cwd() while the new code resolves against plan.repoRoot) is pre-existing and not reachable through the CLI's same-cwd invocation; left for a follow-up issue rather than widened into this PR. Verified: lint:ci exit 0; full suite 27/27 chunks, 37,739 tests, 0 failures, against the true merge-base; tests/worktree-safety.test.cjs 425/425. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01S4mJZpSNwoyVtRVUQoijfL * test(#4415): make the real-git rows platform-correct The Windows conformance shard caught both rows on their first push, and both failures were mine, not the code's. Path separators: git reports porcelain paths with FORWARD slashes on every platform, while `path.join` yields backslashes on win32, so `includes()` compared separator styles rather than paths and the registration assertions failed. Both sides are normalised before comparison now. Premise setup: the unreadable-worktree row establishes "git cannot traverse the parent" with mode 000, which win32 does not honour for directory traversal at all — the row would have asserted `prunable` against a perfectly readable worktree and failed for a reason unrelated to the behaviour under test. It now skips on win32 for the same reason it already skipped as root, with both reasons stated together. Verified: lint:ci exit 0; tests/worktree-safety.test.cjs 425/425 locally. The Windows shard is the real check for the separator fix, since macOS cannot reproduce it. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01S4mJZpSNwoyVtRVUQoijfL * fix(#4415): warn when an entry is accepted as absent, giving prunable its consumer Maintainer review round 3, both Medium findings — they close together, as the review noted. The absent path reported `merged_removed`/`ok` indistinguishably from an ordinary merge. This code cannot tell "the harness cleanly removed a finished executor" from "an operator or an external process removed this path": git keeps the path -> branch registration and `statSync` reports ENOENT in both cases. Before this path existed every anomalous absence blocked loudly, so accepting the routine case silently took the operator's only signal away from the case that is not routine. The module already carries an advisory channel for a materially less risky condition — scope conformance, a few lines below — so withholding one here was inconsistent with its own pattern. `WAVE_CLEANUP_WARNING.ACCEPTED_ABSENT_WORKTREE` is now emitted at both acceptance sites, carrying git's own `prunable` reason. Advisory, never a gate: the entry still merges. That also gives `WorktreeEntry.prunable` a consumer. It was parsed, documented as "worth surfacing to an operator", and then never read — the errno rework made it unused for the predicate and the parsing stayed behind. Quoting git's reason here is what it was for. The bare-marker test was vacuous, as the review said: it asserted `merged_removed`, which is driven by `confirmedGone` and the branch match, not by the bare-marker parsing it claimed to cover, so a regression in that parsing would not have reddened it. It now asserts the parsed value reaches the warning. A bare `prunable` line normalises to the literal 'prunable' — a truthiness signal, not a reason — so the warning reports null there rather than quoting a marker back at an operator as though git had said something. `WAVE_CLEANUP_WARNING`'s locked code set is updated deliberately, with the reason recorded in the test: the lock exists so a new advisory code is a decision rather than something that appears because a branch needed one. Verified: mutation — suppressing the warning at both sites turns both new rows red; lint:ci exit 0; full suite 27/27 chunks, 38,245 tests, 0 failures; tests/worktree-safety.test.cjs 426/426. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01S4mJZpSNwoyVtRVUQoijfL --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com> Co-authored-by: Tom Boucher <trekkie@nomorestars.com>	2026-09-20 04:00:45 -04:00
Tom Boucher	822934c901	fix(#4794): decision-coverage answers an unmeasured shape on could-not-parse; a non-file context path fails closed (#4889) ... * test(#4794): failing-first — could-not-parse must answer an unmeasured shape; a directory context path fails closed * fix(#4794): could-not-parse answers an unmeasured shape (null counts, unreadable ids, no uncovered); a non-file context path fails closed * chore(#4794): backfill changeset PR number (4889) * test(#4794): skip the directory-identity probe when the platform cannot discriminate (windows runner volume collapse, measured) Two consecutive windows conformance runs failed the probe with measured identical (dev, ino) for two distinct mkdtemp directories (dev=3606225537, ino=9007199255243448 for both) — a runner-volume property, not a regression in the guard. On such a platform the guard's identity containment degrades to refuse-everything (fail-closed, documented); the probe asserts capability, so the honest response is an explicit t.skip carrying the measurement (ADR-2719 §6), not a red lane for every PR. --------- Co-authored-by: sim <sim@local>	2026-09-20 02:53:58 -04:00
0xdhx	8a5166598c	fix(#4830): re-land #4768's letter-suffix phase-id fix and its lint-phase-id-drift ratchets on current next (#4873) ... * fix(#4830): re-land #4768's letter-suffix phase-id fix and its lint-phase-id-drift ratchets on current next Commit 740ba0d8a (#4781) removed every change #4768 had merged for #4748: the first-non-digit split at execute-phase.md's two arithmetic sites, the init-emitted `padded_phase` the REVIEW.md lookup binds instead of `printf "%02d"`, the canonical-grammar extractions in autonomous.md and plan-review-convergence.md, the `.changeset/zesty-wolves-tumble.md` fragment, and the three lint-phase-id-drift ratchets with their tests. The guard and the code it guarded left together, so nothing went red. This is a cherry-pick of 092d9256b onto current `next`, resolved against the #4683 threat-id fields on execute-phase.md's Parse-JSON line, with the changeset `pr:` reset to the placeholder and the compact-content benchmark baseline regenerated against the current base. (cherry picked from commit 092d9256b8) Emitted-Drift-Ack-Growth: autonomous.md — restores #4768's canonical-grammar extraction and its explanatory comment for --from/--to/--only Emitted-Drift-Ack-Growth: execute-phase.md — restores #4768's first-non-digit split at two arithmetic sites and the padded_phase binding for the REVIEW.md lookup Emitted-Drift-Ack-Growth: plan-review-convergence.md — restores #4768's canonical-grammar phase extraction and its comment Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AX7LXxc3uAkGki6iaYiAMP * chore(#4830): set changeset fragment pr to 4873 --------- Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com> Co-authored-by: Tom Boucher <trekkie@nomorestars.com>	2026-09-20 00:01:21 -04:00
Tom Boucher	09e1110e68	fix(#4788): code spans in the decision bold lead-in are opaque to the separator grammar (#4883) ... * test(#4788): failing-first — code spans in the decision bold lead-in are opaque to the separator grammar * fix(#4788): the decision lead-in runs are code-span-aware — a backticked span is data, never grammar * chore(#4788): backfill changeset PR number (4881) * chore(#4788): correct changeset PR number (4883, was a guessed 4881) --------- Co-authored-by: sim <sim@local>	2026-09-19 22:46:48 -04:00
Tom Boucher	5906a24ede	fix(#4786): plan-row detection accepts the bare planId stem — suffix-less hand-written lists tick in place (#4880) ... * test(#4786): failing-first — a suffix-less hand-written plan list is ticked in place, never duplicated * fix(#4786): plan-row detection accepts the bare planId stem — a suffix-less hand-written list is ticked in place, never duplicated * chore(#4786): backfill changeset PR number (4880) --------- Co-authored-by: sim <sim@local>	2026-09-19 21:06:05 -04:00
Tom Boucher	d435723c95	fix(#4782): claude's agents kind skips compact variants — consumed only by the non-claude persona gate (#4878) ... * test(#4782): failing-first — claude install must not stage compact agent variants (and must prune stale ones) * fix(#4782): claude's agents kind skips compact variants — they are consumed only by the non-claude persona-fallback gate Emitted-Drift-Ack-Hash: agents/gsd-advisor-researcher.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-ai-researcher.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-assumptions-analyzer.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-code-fixer.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-code-reviewer.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-codebase-mapper.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-debug-session-manager.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-doc-classifier.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-doc-synthesizer.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-doc-verifier.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-doc-writer.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-dom-verifier.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-domain-researcher.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-eval-auditor.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-eval-planner.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-framework-selector.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-integration-checker.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-intel-updater.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-mempalace-curator.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-nyquist-auditor.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-pattern-mapper.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-project-researcher.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-research-synthesizer.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-roadmapper.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-security-auditor.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-ui-auditor.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-ui-checker.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-ui-researcher.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies Emitted-Drift-Ack-Hash: agents/gsd-user-profiler.compact.md — #4782: claude never selects compact (the init agent-skills persona gate is non-claude-only), so the claude agents kind no longer stages name-shadowing copies * chore(#4782): changeset fragment * chore(#4782): backfill changeset PR number (4878) --------- Co-authored-by: sim <sim@local>	2026-09-19 17:25:21 -04:00
Tom Boucher	4fe2837ab5	fix(#4774): plan-criteria R4 requires the pipe not to be doubled — a logical-OR fallback is handled, not swallowed (#4877) ... * test(#4774): failing-first — R4 must not read a logical-OR fallback as a pipeline stage * fix(#4774): R4 requires the pipe not to be doubled — a logical-OR fallback is a handled failure, not a swallowed one Also corrects the rows-3/4 test's makeCriteriaPlan usage (second arg is the <verify> block, not a second criteria line). * chore(#4774): backfill changeset PR number (4877) --------- Co-authored-by: sim <sim@local>	2026-09-19 15:38:16 -04:00
Tom Boucher	a87b83d485	fix(#4764): dep_phases extracts only Phase-prefixed references from Depends-on prose (#4876) ... * test(#4764): failing-first — dep_phases must extract only Phase-prefixed references, never dates/shas/ledger ids/self * fix(#4764): dep_phases anchors phase references to their 'Phase' prose context and never emits the row's own number * fix(#4764): review fold-ins — hoist the anchored dep-reference grammar to phase-id, cover Oxford lists and hyphen ranges, repair the property test Adversarial review found: Oxford-comma lists under-extracted ('Phases 1, 2, and 3' dropped the tail member — a silent real-blocker clear, the dangerous direction); hyphen ranges ('Phases 1-3') kept only the first endpoint; the property test called fc.hexaString (absent in fast-check 4.8, threw every run) and passed the junk arbitrary unspread (vacuous guard) with no completeness assertion; planning-inspect's extractDependencyTokens carried the same whole-field scrape (generative-fix divergence). The anchored grammar now lives beside PHASE_NUMBER_TOKEN_SOURCE in phase-id.cts and both readers interpolate it. * chore(#4764): backfill changeset PR number (4876) --------- Co-authored-by: sim <sim@local>	2026-09-19 13:07:03 -04:00
Tom Boucher	d36514b816	fix(#4758): rescue resolves a relative worktree_path against repoRoot, not process.cwd() (#4872) ... * test(#4758): failing-first — rescue must resolve a relative worktree_path against repoRoot * fix(#4758): rescue resolves a relative worktree_path against repoRoot, not process.cwd() * test(#4758): review fold-ins — t.after cleanup pattern, post-resolution reader contract comment * chore(#4758): changeset fragment * chore(#4758): backfill changeset PR number (4872) * test(#4758): windows lanes key rescue fakes on resolved path identity, not verbatim strings win32 path.resolve rewrites driveless-absolute POSIX-style fixture values to the current drive, so the rescue's (correct) resolved-path handoff stopped matching verbatim string keys: #3804/#245/#2556/B7/#2852 fakes silently skipped the rescue and my seam test compared against a POSIX literal. Fakes now key on path.resolve(repoRoot, …) identity — the same semantics the code and git -C use — so every rescue test exercises the rescue on every platform. --------- Co-authored-by: sim <sim@local>	2026-09-19 09:41:51 -04:00
Tom Boucher	63edc777e6	fix(#4588): observe fork-from-HEAD from prior harness worktrees before degrading (#4868) ... * test(#4588): observed fork-from-HEAD must suppress the stale-origin degrade (failing first) * fix(#4588): observe fork-from-HEAD from prior harness worktrees before degrading * fix(#4588): a throwing probe git call is an inconclusive observation, not a crash * test(#4588): name the observed reason in the inconclusive-row assertions * test(#4588): hermetic state I/O in the inconclusive-row fixtures * chore(#4588): changeset fragment * test(#4588): contained cache I/O, emit-payload assertions (review fold-ins) * fix(#4588): review fold-ins — invalidation fixture, hermetic confirm row, gate comment, cache+scope rows * chore(#4588): backfill changeset PR number (4868) --------- Co-authored-by: sim <sim@local>	2026-09-18 23:41:44 -04:00
Tom Boucher	9a41a95212	fix(#4717): consult the per-install runtime marker at both identity seams (#4861) ... * test(#4717): add failing-first coverage for the two runtime-identity marker seams * fix(#4717): consult the per-install runtime marker at both identity seams resolveReportedRuntime (agent_runtime) and loadConfigResolved (config.runtime) both ignored the per-install .gsd-runtime marker that resolveRuntime and the model-resolver gate already read. On a multi-runtime machine (e.g. a globally exported CODEX_HOME), host sniffing misreported every Claude Code session as codex, and a shared defaults.json stamped by the first non-Claude install leaked its runtime to every other one. Seam 1: the reported-runtime ladder becomes explicit > install marker > host detection > claude. Seam 2: loadConfigResolved fills an empty config.runtime from GSD_RUNTIME then the marker, copy-on-write (the builtin-defaults branch returns a shared object). Explicit runtimes and marker-less trees are unchanged. * fix(#4717): a marker-detected runtime opts into its tier map (decision a) * fix(#4717): stamped-defaults leg, marker fail-safe, docs, review fold-ins * chore(#4717): backfill changeset PR number (4861) --------- Co-authored-by: sim <sim@local>	2026-09-18 13:50:01 -04:00
Tom Boucher	e1f72cd324	fix(#4741): the plan checkbox tick respects the superseded exclusion (#4851) ... * test(#4741): a superseded plan must not be ticked from its summary (failing first) * fix(#4741): the plan checkbox tick respects the superseded-plan exclusion * fix(#4741): review fold-ins — changeset typo, dedupe planId derivation * test(#4741): exercise a halted summary on the active plan in the #2830 pin * chore(#4741): backfill changeset PR number (4851) --------- Co-authored-by: sim <sim@local>	2026-09-18 06:33:30 -04:00
Tom Boucher	c5629bbe74	fix(#4734): degrade worktree isolation when the root has no git repository (#4843) ... * test(#4734): non-git root must degrade worktree isolation (failing first) * fix(#4734): degrade worktree isolation when the root has no git repository * fix(#4734): review fold-ins — 3972 ladder fixture, parity fixture, docs row, message wording * chore(#4734): backfill changeset PR number (4843) --------- Co-authored-by: sim <sim@local>	2026-09-18 03:16:25 -04:00
Tom Boucher	8d0b6868ae	fix(#4725): write normalization preserves tight paragraph-list shape (#4842) ... * test(#4725): write normalization must not reflow untouched prose (failing first) * fix(#4725): stop write normalization injecting a blank before a list after prose * test(#4725): repair ordered-list fixture and list-spacing snapshot * test(#4725): assert whole-file prose stability, fix heading-list comment * chore(#4725): backfill changeset PR number (4842) --------- Co-authored-by: sim <sim@local>	2026-09-18 01:38:14 -04:00
Tom Boucher	c9a5cc3e12	fix(#4683): detect cross-plan threat-ID duplicates before execution (#4828) ... admin_reason: missing-secondary-reviewer — self-authored overnight sweep; two orthogonal agent reviews ran (isolated adversarial REQUEST-CHANGES with all six findings dispositioned, plus a bypass/consumer-lens APPROVE) and the sha-pinned bench passed 46362/0 on the merged head.	2026-09-17 19:06:55 -04:00
Tom Boucher	bff99a8bb5	fix(#4731): read hard-wrapped Goal/Requirements fields past the line break (#4826) ... admin_reason: missing-secondary-reviewer — self-authored overnight sweep; isolated adversarial review round completed (MEDIUM table-bleed finding fixed with RED/GREEN evidence) and sha-pinned bench 46331/0 on the merged head.	2026-09-17 15:59:38 -04:00
Tom Boucher	fb3e228a0d	fix(#4724): classify Surefire/Failsafe XML as RED evidence (#4825) ... * test(#4724): add failing-first coverage for Surefire XML RED evidence * fix(#4724): classify Surefire/Failsafe XML as RED evidence check tdd-red-evidence parsed only node:test TAP, so a JVM project's genuine Maven red scored INVALID_RED while hand-written synthetic TAP scored RED_EVIDENCE_OK — the gate was passable only by fabricating its input (issue #4724's measured repro). classifyRedEvidence detects Surefire/Failsafe XML (a <testsuite> element) and parses it by TAG-BOUNDARY scanning: each <testcase> owns its own tag (self-closing) or the segment up to its </testcase> closer, so the issue's warned-about spanning trap (a lazy lazy match from a green self-closing case to the next closing tag) cannot misreport names. A <failure> or <error> child marks the case failing; the target matches at class granularity (exact classname, dotted-suffix, or method name). Any parse anomaly degrades to not-failing — the module stays fail-closed and PURE (no fs/clock; report freshness remains the workflow's run-start check per the issue's implementation notes). TAP classification is byte-identical: all existing fixtures stay green. * test(#4724): pin the scanner hardening — truncation, TAP-message flip, CDATA phantom * docs(#4724): backfill changeset PR number --------- Co-authored-by: sim <sim@local>	2026-09-17 11:30:23 -04:00
Tom Boucher	7d0c6339d0	fix(#4705): emit Antigravity-native tool names as a YAML sequence (#4822) ... * test(#4705): add failing-first coverage for native Antigravity tool sequences * fix(#4705): emit Antigravity-native tool names as a YAML sequence convertClaudeAgentToAntigravityAgent and the installer's twin emitted Gemini CLI tool names as a comma-separated scalar. Antigravity's documented subagent contract (antigravity.google/docs/subagents) wants a YAML sequence of native names — view_file, grep_search, run_command, replace_file_content are the documented examples, and wrong or malformed grants can hang the subagent per Antigravity's own warning. Map values move to the native vocabulary where documented (Read -> view_file, Edit -> replace_file_content, Bash -> run_command, Grep -> grep_search); undocumented entries keep their best-known grant rather than being dropped (dropping would silently remove a restriction). The emitter writes one '- name' item per line; an agent whose every tool was filtered emits an explicit tools: [] instead of an empty scalar. Pre-existing pins updated to the native vocabulary. * test(#4705): update the #4727 map-value pin to the Antigravity-native vocabulary The #4727-era pin held the map VALUES at the Gemini CLI dialect on the belief that Antigravity speaks it; the confirmed bug #4705 (with Antigravity's own documented subagent contract) supersedes that for the four documented names. Key/shape pinning is preserved; only the values move. * docs(#4705): backfill changeset PR number --------- Co-authored-by: sim <sim@local>	2026-09-17 08:12:18 -04:00
Tom Boucher	d707318e0c	fix(#4699): skip already-complete phases in the next_phase cascade (#4820) ... * test(#4699): add failing-first coverage for skipping complete phases in next_phase * fix(#4699): skip already-complete phases in the next_phase cascade Both next-phase scans selected the numerically lowest phase above N without consulting completion state, so completing a reopened phase persisted an already-[x] phase as STATE.md current_phase while roadmap.analyze correctly named the outstanding one (issue repro: completing 2 with phases 1 and 3 already [x] returned next_phase 03). The cascade collects the complete phase numbers from the roadmap checkboxes (milestone-scoped, comparePhaseNum-deduped) and skips them in both the disk scan and the roadmap scan; a [x] checkbox row and its heading sibling both name a phase that is never next. Heading-only and checkbox-less roadmaps behave exactly as before. * test(#4699): align the negative-control expectation with the disk spelling * test(#4699): pin the STATE.md persistence and the all-later-complete tail corner Review findings: the regression never asserted STATE.md current_phase (the issue's actual harm), and the all-later-phases-[x] corner (is_last_phase true, next_phase null) was unpinned. A changeset fragment is included. * docs(#4699): backfill changeset PR number --------- Co-authored-by: sim <sim@local>	2026-09-17 04:17:39 -04:00
Tom Boucher	2bfff17ff8	fix(#4682): route stale verification to the verifier regeneration path (#4818) ... * test(#4682): add failing-first coverage for stale verification routing * fix(#4682): route stale verification to the verifier regeneration path The stale routing entry sent users to /gsd-verify-work — but verify-work never rewrites VERIFICATION.md (its only write is the human_needed canonicalization), so following the advice re-ran UAT, reached the same stale check, and looped. init's projector and execute-phase's generic next_command presentation both mirror this entry, so the dead end appeared on three surfaces. The stale entry now routes to execute-phase, and execute-phase's all-plans-complete resume tree gains a stale arm (as a steps/ part, keeping the spine under its frozen ADR-857 ceiling) mirroring the missing route: skip cross_ai_delegation/execute_waves/checkpoint_handling, continue at aggregate_results, and let verify_phase_goal re-dispatch the gsd-verifier — regenerating VERIFICATION.md and its digest, marked phase or not. The non-stale fall-through. Staleness detection, the digest format (#4623), every other routing entry, and the #3684 resume arms are untouched. Emitted-Drift-Ack-Growth: verify-work.md — stale stop rewritten to dispatch the verifier and re-check (#4682) Emitted-Drift-Ack-Growth: execute-phase.md — VERIFY_STATUS == stale resume arm added to condition 3 (#4682) * test(#4682): register the stale-reverification part and align projected commands The new steps/ part must be registered in the inventory manifest and the per-runtime golden install trees (regen:derived); the projected stale next_command is /gsd-execute-phase <phase> (formatGsdSlash prefixes the runtime surface), the human_needed bare-report probe keeps routing to verify-work (unchanged semantics), and init-manager's recommended action follows the new command. * test(#4682): prefix the remaining stale routing assertions with the runtime surface Nine stale next_command assertions and the human_needed bare-report probe still carried the unprefixed or flipped forms from the earlier line-number edit; all now assert the shipped /gsd-execute-phase <phase> projection, with the human_needed probe reverted to its unchanged verify-work routing. * test(#4682): align the last stale projection assertions with the execute-phase route * docs(#4682): backfill changeset PR number * test(#4682): refresh the compact-content baseline after the rebase The rebase onto the #4670 squash brought verify-work.md's bounded reconciliation text into this branch; the committed compact-content baseline now reflects the post-rebase split sizes. Local --check is clean; the previous bench drift (+243) was the baseline, not the diff. * fix(#4682): carry the response_language directive in the stale-reverification part The new steps/ part is its own coverage unit for lint-response-language-coverage; it takes the shared canonical directive line like its sibling execute-phase parts. --------- Co-authored-by: sim <sim@local>	2026-09-17 02:23:03 -04:00
Tom Boucher	85545a77a5	fix(#4663): gate the canonicalization on the uat-passed predicate (#4809) ... * test(#4663): add failing-first contract coverage for the blocked-uat canonicalization gate verify-work.md's complete_session step flips VERIFICATION.md to passed on 'zero issues' alone, so a session whose every UAT row is blocked (a session that observed nothing) canonicalizes the report. Pins the deployed contract the fix must satisfy: the flip runs the unflagged phase uat-passed predicate inside the human_needed branch, frontmatter.set sits inside a passed==true guard, a refusal message carries the blocker count and keeps human_needed, and an indeterminate pre-check fails closed. All four new assertions are RED until the workflow grows the guard. * fix(#4663): gate the canonicalization on the uat-passed predicate complete_session flipped VERIFICATION.md to passed whenever the session recorded zero issues and the status was human_needed — but blocked rows are not issues by this workflow's own rule, so a 0-passed / 0-issues / N-blocked session (one that observed nothing) rewrote the canonical report to passed. Every later reader (transition.md's preliminary check, resume paths, validate-phase, verification.status) then inherited the unearned pass while the phase-close predicate correctly refused it. The flip now runs the phase-close predicate in a new --uat-only form before canonicalizing: UAT rows evaluated (at least one pass, no pending/blocked/failed/unexplained-skip row), VERIFICATION-status blockers skipped — they must be, because the report still reads human_needed at pre-check time and that status is itself a blocking verification entry, so the full predicate could never pass there and the flip would deadlock (found by isolated review, probed). The --require-verification call stays the transition gate; the refusal branch reports the blocker count and keeps human_needed; an indeterminate pre-check fails closed. Emitted-Drift-Ack-Growth: verify-work.md — canonicalize block gains the uat-only pre-check and refusal branch (#4663) * test(#4663): align the canonicalize pre-check needles with the shipped line The workflow line carries a 2>/dev/null redirect the needles did not include, so both pre-check assertions fail against the committed fix (fixed-string grep verified). Reviewer-found; needle and message aligned. * fix(#4663): reword the canonicalize prose and refresh its size baseline The rationale paragraph mentioned the flagged transition-gate call by its flag, putting a --require-verification literal before the first phase uat-passed occurrence and breaking the existing ordering pin; the prose now describes it without the literal. verify-work.md's growth also drifted the committed compact-content baseline; regenerated via benchmark-compact-content.cjs --write (derived artifact, report-not-gate contract). Emitted-Drift-Ack-Growth: verify-work.md — canonicalize block gains the uat-only pre-check and refusal branch (#4663) * docs(#4663): backfill changeset PR number --------- Co-authored-by: sim <sim@local>	2026-09-16 17:26:24 -04:00
Tom Boucher	c72fb34e9a	fix(#4658): give the ui plan gate's evidence check a native branch (#4807) ... * test(#4658): add failing-first coverage for native frontend evidence hasStaticFrontendEvidence recognised only JS-ecosystem evidence, so computeUiPlanGate could never block for a SwiftUI/Compose/Flutter/XAML project. Adds evidence-level fixtures for the four suggested markers (import-matched for .swift/.kt/.dart, extension-alone for .xaml), the reporter's non-UI Swift control case, marker-exactness and SKIP_DIRS and I/O-degrade negatives, gate-level block assertions through makeProject's new native frontendEvidence modes, and a pinned-seed fast-check property. All new assertions are RED until src/ui-frontend-evidence.cts grows the native branch. * fix(#4658): give the ui plan gate's evidence check a native branch hasStaticFrontendEvidence recognised only JS-ecosystem evidence (a root package.json UI-framework dep, or a .tsx/.jsx/.vue/.svelte file), so computeUiPlanGate could never block for a SwiftUI, Jetpack Compose, Flutter, or .NET MAUI project — the #3312 gate was structurally unreachable for them. Adds a native BFS over the same bounds and skip rules: .xaml is evidence by extension alone (the .tsx analogue), while .swift/.kt/.dart count only when their content carries the ecosystem's UI import marker (import SwiftUI / import UIKit, androidx.compose, package:flutter) — matched on the import, not the extension, so a non-UI Swift package stays silent exactly as the issue's 37-file control case requires. Marker reads are bounded to a 64 KiB prefix; any I/O failure degrades to false per the module contract. The #3718 vocabulary filter, the JS evidence rules, and the weaker-extension exclusion are untouched. * chore(#4658): regenerate the macos conformance tier list The native-evidence additions to tests/check-ui-plan-gate.test.cjs move the file into the macOS conformance tier per the classifier; the committed generated list is a derived artifact and must match the live tests/ tree (the same sync the fragment-single-edit-propagation install test enforces). * fix(#4658): accept both Dart quote styles and extract the shared bounded walk The isolated reviews' remaining findings: the Dart marker carried only the single-quote anchor, missing legal double-quoted imports (a spec-narrowing deviation); the two evidence walks duplicated the subtle MAX_WALK_ENTRIES cap semantics verbatim, so they are extracted into one walkProjectFiles BFS with a visit callback; tests now use the createTempDir helper, shared fixture literals that cannot drift from NATIVE_UI_CONTENT_MARKERS, a double-quoted Flutter import case, and drop a vacuous assertion and a mid-body re-require alias. * docs(#4658): backfill changeset PR number --------- Co-authored-by: sim <sim@local>	2026-09-16 15:29:19 -04:00
Tom Boucher	5e729445d3	fix(#4657): give the ui consideration probe a text_en language channel (#4804) ... * test(#4657): add failing-first coverage for the ui probe's text_en channel Mirrors the #3717/#4156 test shape onto the UI adapter: a failing-first proposeConsiderations regression (Danish text + English text_en must classify as its English equivalent, not land in the #1110 unclassified sentinel), proposeElements/analyzeCoverage/CLI end-to-end pairs, fail-closed text_en validation cases (empty/whitespace/non-string, unconditional under an elements override), a ui-phase.md Step 9.5 workflow-prose contract test, a reference-doc Inputs parity test, and a fast-check property proving any cue-matching prose classifies identically under a cue-free Danish rendering plus text_en. All new assertions are RED until src/ui-consideration-probe.cts and the workflow/reference docs are updated. * fix(#4657): give the ui consideration probe a text_en language channel Element gains an optional text_en; classifyElement's own signature stays untouched (a locked, directly-tested export) and the text_en ?? text selection is pushed to the two classification call sites (proposeConsiderations, proposeElements) instead. text_en is validated fail-closed: an empty or whitespace-only value throws rather than silently winning the ?? fallback and degrading classification to zero kinds. Mirrors #3717/#4156 onto the UI adapter: ui-phase.md Step 9.5 gains the Non-English projects section (mirroring spec-phase Step 5.5) and the ELEMENTS_JSON shape comment documents the field with both zero-applicable guard arms named; the reference doc's Inputs section, the PROBE.ui CONTEXT predicate (with both derived indexes regenerated), and the nav-override test expectation stay in sync. The ui-phase contract test carries the site-scoped allow-test-rule marker and its cluster is registered in the test-file-count allowlist ratchet. Emitted-Drift-Ack-Growth: ui-phase.md — Non-English text_en section, ELEMENTS_JSON shape comment, and two-arm guard wording (#4657) * docs(#4657): backfill changeset PR number --------- Co-authored-by: sim <sim@local>	2026-09-16 13:46:40 -04:00
Tom Boucher	3014775a3f	fix(#4656): expose coverage.unclassified and widen the zero-applicable guards (#4800) ... * fix(#4656): expose coverage.unclassified and widen the zero-applicable guards * fix(#4656): regenerate golden coverage fixtures and update the rollup pin Emitted-Drift-Ack-Growth: spec-phase.md — #4656: guard widened to the all-unclassified case, doc claim corrected Emitted-Drift-Ack-Growth: ui-phase.md — #4656: guard widened identically * fix(#4656): sync edge-probe doc blocks and coverage pins with the new field * fix(#4656): key the mandatory confirmation on the widened guard * docs(#4656): backfill changeset PR number --------- Co-authored-by: sim <sim@local>	2026-09-16 10:33:32 -04:00
0xdhx	003d982c83	fix(#4623): keep repo-wide planning docs out of the verification digest, and accept --files on verification.fingerprint (#4749) ... * fix(#4623): keep repo-wide planning docs out of the verification digest, and accept --files on verification.fingerprint Two defects in the covered-input fingerprint (#4155), one issue. 1. `computeCoveredDigest` hashed the whole bytes of every declared path uniformly, so `.planning/ROADMAP.md` and `.planning/REQUIREMENTS.md` — which every phase rewrites as ordinary bookkeeping, and which the closing phase's own `phase.complete` / `requirements mark-complete` rewrite AFTER the verifier ran — flipped every phase that declared them to `stale` on zero implementation change, and from there `isPhaseComplete` → `init.manager` → `complete-milestone`'s `ALL_PHASES_VERIFIED` gate. Fingerprint v2 leaves any direct child of a planning root out of the hash: `.planning/` itself, plus the phase's own planning root (the parent of its `phases/`, so `planningDir`'s `<project>/` and `workstreams/<ws>/` layouts are covered without the digest knowing what a workstream is — `sharedPlanningRoots` / `isSharedPlanningDoc`, defined by position rather than a name list so the set cannot drift; a root is accepted only when the phase dir sits under a `phases/` directory inside `.planning/`). Such a path is still validated exactly as every other covered path (confined, present, a regular file — the fail-closed contract is unchanged); only its bytes are ignored, and a declaration made only of shared documents fails closed like an empty one. A stored digest names its version, and `readVerificationStatus` now recomputes under THAT version (`parseFingerprintVersion`, `KNOWN_FINGERPRINT_VERSIONS`): a legacy v1 report keeps v1 semantics until it is re-fingerprinted, so the upgrade alone stales nothing; a version this build cannot recompute fails closed. 2. `verification.fingerprint` received a raw positional slice, so `--files a`, `--files "a,b"` and `--files a --files b` all put the literal token into the covered set and failed closed as "a covered file is missing, unreadable, or escapes the project root" — the message that convinced the reporting project the digest was permanently unrecomputable. `parseFingerprintFileArgs` accepts every form (plus `--files=a,b`, freely mixed with bare positionals), treats any other `--flag` and an empty `--files` value as usage errors that say so, and the phase-dir argument must now be an existing directory: omitting it used to take the first covered file as the phase dir and print a plausible digest over the rest at exit 0. Regression tests (tests/verification-status.test.cjs, #4623 block): the cross-phase case from the report, the same-phase `requirements mark-complete` / `phase.complete` cases from the thread, a workstream-scoped root, v1-preserved / unknown-version-stale, the fail-closed cases (missing, directory, escaping symlink, all-shared), every `--files` form against the bare form, the unknown-flag / empty-value / omitted-phase-dir errors, and AC5's zero-file error. Verified failing against the pre-fix source: 29 of 34 fail, the 7 that pass pin behaviour the fix must leave unchanged. Docs: CONTEXT.md Verification Module, agents/gsd-verifier.md's covered_files instruction (rewritten in place — the file sits 21 bytes under its LARGE hard cap), gsd-core/templates/verification-report.md. Fixes #4623 Emitted-Drift-Ack-Growth: gsd-verifier.md — the #4155 covered_files instruction now states that planning-root docs are digest-inert (#4623); +18 bytes, under the LARGE cap Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DCMY8P8s6dp4g3Rxu3nNAi * chore(#4623): set changeset fragment pr to 4749 --------- Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com> Co-authored-by: Tom Boucher <trekkie@nomorestars.com>	2026-09-16 05:26:44 -04:00
Dennis Alexis Valin Dittrich	ad1477d659	enhance(#4154): validate configured entrypoints before reporting install success (#4249) ... * test(260903-m7p): expose configured-entrypoint validation gap * enhance(260903-m7p): validate configured entrypoints before success * test(260903-m7p): require pre-success entrypoint validation * enhance(260903-m7p): gate install success on entrypoints * test(260903-m7p): cover configured entrypoints across runtimes * enhance(260903-m7p): cover emitted runtime entrypoints * fix(260903-m7p): sandbox HOME in finishInstall test and fix changeset pr number - finishInstall(...'cline'...) calls writeNonClaudeDefaults(runtime) in-process before the new configured-entrypoint assertion throws. Without a HOME + config-location-env sandbox that write resolved through the ambient environment and landed in the developer's live ~/.gsd (confirmed absent on origin/next baseline, present only on this branch — full-suite HERMETICITY WARNING). Sandbox HOME/USERPROFILE and scrub config-location env for the duration of the test, matching the existing in-process finishInstall/ install() pattern in tests/install.test.cjs (#2665). - .changeset/quick-wasps-sing.md: pr: 0 is a never-backfilled placeholder (CONTRIBUTING.md) that fails changeset-lint's invalid_pr check; set to the fork PR number until the upstream PR number is known. * fix(260903-m7p): repair cross-platform and pre-existing shape fallout - tests/configured-entrypoint-validation.test.cjs: the win32 branch of ensureCodexHooksJsonSessionStart writes a .cmd shim under <codexRoot>/hooks/; create that dir in the test (the real installer only calls this once hooks/gsd-check-update.js already exists) and assert the platform-common entrypoint shape instead of a fixed non-Windows array, since win32 legitimately emits two entries (cmd shim + script). - tests/install.test.cjs: finishInstall's shared settings-json return now carries configuredEntrypoints/rollbackInstallerMigrations for every runtime on that path (trae included, not just Claude/Cursor/Windsurf); update the trae install() exact-shape assertion to match. * fix(260903-m7p): keep .sh interpreter tracking consistent with unresolved bash configuredEntrypointsForHook's shell branch dropped interpreterCandidates entirely when resolveBashExecutable returned null, unlike the sibling portableHooks runner entry a few lines below (which correctly falls back to the literal 'bash' token). Found via agy adversarial review; verified unreachable through the current call graph (buildHookCommand's own resolveBashRunner==null gate already short-circuits before recordConfiguredHookCommand runs), so this is a defensive consistency fix, not a live-bug patch — kept for the next caller that does not share that gate. * chore(260903-m7p): backfill changeset pr number to the opened upstream PR .changeset/quick-wasps-sing.md carried the fork PR number (16) as a placeholder until the upstream PR existed; open-gsd/gsd-core#4249 is now open, so record its real number per CONTRIBUTING.md's changeset pr-field convention. * fix(#4154): track already-registered hooks for entrypoint validation on update applySettingsJsonHooks registers each guard hook only if absent, so a hook already present from a prior install keeps its stale on-disk command. The new entrypoint tracker always records the freshly-computed command for it, which never matches what is actually persisted, so the exact-string filter in finishInstall silently dropped it from validation — the Blocker case this feature exists to catch (an already-installed entrypoint going stale between installs) was exactly the case it never validated. Match on the managed script's basename instead, which the persisted command carries either way, so an already-registered hook stays in the validated set. Regression test forces this path by mutating a freshly-installed hook's persisted command before a second install. * fix(#4154): distinguish an unreadable script from a missing one validateConfiguredEntrypoints folded an EACCES statSync failure into the same 'missing' reason as ENOENT, misreporting a real permission problem as an absent file. Check the error code and report 'unreadable' instead. * docs(#4154): document entrypoint validation's rollback and PATH scope CONTEXT.md's Runtime Hooks Surface Module / Installer Module entries had no mention of ConfiguredEntrypoint/validateConfiguredEntrypoints, despite bin/install.js x CONTEXT.md being this repo's strongest co-change pairing. The update-gsd.md how-to overstated what a validation failure undoes: for Codex/Cursor/Windsurf/Kimi, their own writer already persisted hooks.json/ config.toml inside install() before the aggregate validation call runs, so there is no rollback path for that write regardless of "where available" phrasing. Also note that interpreter resolution checks the installer's own PATH, not necessarily the PATH a hook fires under later (#2979 launchers). * chore(#4154): point changeset pr field at the fork PR while CI runs there Mirrors the branch's own prior backfill commit: pr: matches whichever PR number changeset-lint is currently validating against (fork PR #16 during the fork-first CI/review loop), flipped back to the upstream PR number right before the final push to open-gsd/gsd-core. * fix(#4249): address adversarial-review findings in entrypoint validation An internal adversarial review (agy/gemini-3.8-flash-high) of the whole PR found several real gaps beyond the human reviewer's Blocker, verified against source before fixing: - Codex's install() result bound rollbackInstallerMigrations to the narrow installer-migrations-only rollback instead of restoreCodexSnapshot (#3245), the full pre-install snapshot/restore Codex already owns for exactly this case — a validation failure discovered outside install() reverted nothing of the config.toml/hooks.json that call had already written. - The register-only-if-absent basename match from the prior fix used a bare substring, which an unrelated user command mentioning the same filename could false-positive into GSD's validated set — anchored on the `/hooks/<basename>` path segment instead. - nodeCandidates checked raw process.execPath (always true — we're running in that process) instead of normalizeNodePath's stable version-manager alias, the same one buildNodeRunnerChainToken bakes as its first choice — a false green regardless of whether that alias itself still resolves. - An entry with no interpreterCandidates (Cline's PreToolUse hook, or a Windows-Claude .sh hook invoked without a bash runner) runs via its own shebang; validateConfiguredEntrypoints checked only file-type, never the execute bit. Cline's writer also never reported an entrypoint at all. - Duplicate (configPath, scriptPath) entries (e.g. Kimi's context-monitor hook registered across several events) were validated once per duplicate. Each fix is covered by a new or extended test; the Codex one required inlining runCodexInstall's env sandboxing so the rollback closure — which re-resolves the $HOME-relative skills root live — runs before the sandbox is torn down, matching how installAllRuntimes' real aggregate gate calls it. * docs(#4249): document the round-2 entrypoint-validation fixes Runtime Hooks Surface Module and Installer Module entries now name ConfiguredEntrypoint's not-executable reason, the normalizeNodePath alignment, Cline's tracked hook, and which install() result the finishInstall/installAllRuntimes rollback path actually reverts per runtime (Codex's full snapshot vs. the others' narrow migrations-only rollback). * chore(#4249): point changeset pr field at the upstream PR now that fork CI is green * fix(#4249): address agy adversarial-review findings - validateConfiguredEntrypoints: statSync alone never detects a chmod-000 script (it only needs parent-dir search permission), so an interpreter-invoked entry with an unreadable script passed validation. Add an explicit R_OK check for the interpreterCandidates branch only — the candidate-less/shebang branch already has its own X_OK gate. - docs/how-to/update-gsd.md: the blanket "does not revert" claim was false for Codex, which reverts config.toml/hooks.json via its full pre-install snapshot; qualify it per runtime. - tests/codex-config.test.cjs: the #4249 rollback regression test asserted skills/ and VERSION were reverted but never asserted config.toml/hooks.json were too, despite the test's own stated intent. - CONTEXT.md: qualify which interpreterCandidates entries get normalizeNodePath'd (Node hooks only, not .sh/bash) and note Codex's Windows .cmd shim as a third candidate-less case that relies on extension dispatch, not a shebang. * fix(#4249): validate Cline's PATH-dependent interpreter, not just its execute bit Cline's hook is a hybrid: it self-executes via '#!/usr/bin/env node', so it needs the execute bit (like any shebang-invoked entry), but its interpreter is looked up on PATH by 'env' at hook-fire time (unlike every other GSD JS hook, which bakes an absolute node path specifically to avoid that dependency). The candidate-less/interpreterCandidates fork treated these as mutually exclusive, so Cline's entry silently skipped interpreter resolution entirely — a completely missing 'node' on PATH would still validate successfully. Add an orthogonal selfExecutable flag so both checks run for entries that need them. (CodeRabbit finding on the fork rehearsal PR.) * fix(#4249): address second-round adversarial review findings (opus + agy) - validateConfiguredEntrypoints: R_OK now runs for every scriptOk entry, not just interpreterCandidates ones — a self-executable shebang script is still opened and read by its kernel-invoked interpreter, so X_OK alone never proved it was readable. - selfExecutable is now the sole, explicit source of truth for the execute-bit check (every producer that needs it sets the flag) instead of being partly inferred from an absent interpreterCandidates, which Cline's hybrid entry also carries. - The execute-bit check now skips explicitly on win32 (matching resolveExecutableBinary's own carve-out) instead of relying on Node's accessSync(X_OK)-as-F_OK no-op, which only protects a real Windows machine and not a test that simulates win32 on a POSIX runner. - bin/install.js: fixed a stale comment claiming no runtime's install()-time writes have a rollback path — Codex's does (restoreCodexSnapshot) — and added the omitted Cline to both that comment and CONTEXT.md's equivalent lists. - CONTEXT.md: fixed the Cline description left stale by the previous commit's selfExecutable addition, and rewrote the validation-mechanism paragraph for clarity (writing-for-agents pass). - docs/how-to/update-gsd.md: split an overloaded 4-clause sentence. - Removed a fault-injection integration test that could not reliably exercise the real installAllRuntimes -> finalize -> rollback wiring without fighting the installer's own pre-registration existence guards; the constituent pieces remain covered individually. * fix(#4249): pin platform in X_OK-testing entries so they're deterministic cross-CI-runner X_OK is a POSIX-only concept, skipped entirely when an entry's platform is win32 (matching production). Two test entries omitted platform, defaulting to process.platform — on an actual windows-latest CI runner that silently skipped the very check they were meant to exercise, turning 'not-executable' into a false pass. Pin platform: 'linux' so these are deterministic regardless of which OS runs the suite. * fix(#4249): classify EPERM the same as EACCES in statSync error handling Windows raises EPERM (not EACCES) for a parent directory that couldn't be traversed into — was falling through to 'missing', misreporting a genuine permission problem as a nonexistent path. * docs(#4249): address final CodeRabbit doc-completeness findings - CONTEXT.md: install()'s documented result shape omitted configuredEntrypoints; the ConfiguredEntrypoint shape omitted selfExecutable. - docs/how-to/update-gsd.md: the failure-mode sentence omitted unreadable and lacks-execute-permission, which the installer also rejects. * fix(#4249): stop double-validating every configured entrypoint on install/update installAllRuntimes' finalize() already runs assertConfiguredEntrypoints once over the aggregate set; finishInstall then re-ran the identical check per runtime in the printSummaries loop right after, so every entrypoint paid its statSync/accessSync/interpreter-resolution cost twice on every install and update. Add entrypointsAlreadyValidated to skip the redundant pass specifically on that path, while leaving the check intact for any caller that invokes finishInstall directly. * chore(#4154): point changeset pr field at rehearsal fork PR while CI runs there * perf(#4249): memoize interpreter candidate resolution across entrypoints resolveExecutableBinary walked PATH once per (entry, candidate) pair; a typical install has a dozen-plus entries sharing the same few candidate lists (process.execPath for JS hooks, bash for shell hooks). Cache by (platform, candidate) so each distinct pair resolves once per validation call instead of once per entry. * chore(#4249): point changeset pr field at the rebased rehearsal fork PR * fix(#4249): drop entrypoint tracking from the now-dead Codex event writer #2586 (landed on next after this branch forked) removed install.js's CODEX_EXTENDED_HOOK_EVENTS registration loop, so ensureCodexHooksJsonEvent no longer runs during install or update. The ConfiguredEntrypoint records this branch added inside it were therefore unreachable and untested. Restore the function to its upstream shape; the entrypoints it used to report were never collected by any caller. * refactor(#4249): drop the revalidation bypass flag and the candidate cache Both were this PR's own micro-optimisations over a set of roughly a dozen entries. `entrypointsAlreadyValidated` let a caller turn the finishInstall gate off to save one statSync/accessSync pass; `resolvedCandidateCache` memoised resolveExecutableBinary across entries that are already deduped by (configPath, scriptPath). Neither is measurable, and the flag was the only way to reach finishInstall with validation disabled. finishInstall now always validates what it is given. * chore(#4249): point the changeset pr field back at the upstream PR * refactor(#4249): track settings.json entrypoints without the hooksSurface gate The install-surface writer only tracked configured entrypoints when the runtime's descriptor also declared `hooksSurface: 'settings-json'`. Nothing asserts that axis agrees with `installSurface`, so a descriptor that broke the coupling would silently pass `configuredEntrypoints: undefined` and drop that runtime out of the validation this PR adds — reintroducing the exact 'reports Done! over a broken entrypoint' failure #4154 exists to close. Remove the dependence rather than test it: everything recorded on this path lands in settings.json by construction, and the registered-command filter already discards entries no persisted hook references. * chore(#4249): put the changeset body in the documented two-part format CONTRIBUTING.md and .changeset/README.md both show `**<bold change>** — <symptom-led explanation>.`; the fragment was a single unbolded sentence. * chore(#4249): point the changeset pr field at the rehearsal fork PR while CI runs there * fix(#4249): restore the whole manifest-tracked GSD file set on Codex rollback #3245's snapshot covers config.toml, hooks.json, skills/gsd-*, agents/gsd-* and gsd-core/VERSION. The install overwrites every other GSD-owned file too — hooks/, gsd-core/CHANGELOG.md, scripts/, gsd-core/.gsd-runtime, the manifest itself — before the entrypoint-validation gate runs, so a validation failure left the new payload sitting on top of the restored old config. Snapshot the file set the PREVIOUS install's gsd-file-manifest.json claims, before runInstallerMigrations so the bytes are the true pre-install state, and restore it from both Codex rollback closures ahead of the per-surface restores. Files only the failed install introduced are removed, read from the manifest now on disk. The manifest is already the authoritative record of what GSD owns, so no second hand-written list can drift out of sync, and user-owned files are never snapshotted or removed. Every path is confined through resolveInstallRelativePath, so a hand-edited manifest cannot turn rollback into an arbitrary-path write. Non-Codex runtimes are unaffected: the snapshot is gated on the same tomlConfigInstall + non-minimal condition as #3245's. * fix(#4249): keep the managed-file snapshot honest in minimal mode and on a bad manifest Two follow-on defects in the previous commit's snapshot: - The capture was gated on `!isMinimalMode`, copied from #3245. A core/ --minimal Codex install still writes gsd-core/, hooks/, scripts/ and the manifest, and restoreCodexSnapshot is reachable in that mode (#2695), so the snapshot came back empty while the rollback still ran — and its removal pass would have deleted every file the new manifest lists. Gate on tomlConfigInstall alone, matching where the rollback actually reaches. - An unreadable or unparseable prior manifest was caught alongside ENOENT and treated as a fresh install. That is the same empty-snapshot state, so a failed update over a real install with a corrupt manifest could delete its prior payload. Track whether the pre-install GSD-owned set is KNOWN: ENOENT means known-empty; any other read error or a parse failure means unknown, and the restore closure returns without touching anything, degrading to #3245's narrower rollback. Deliberately not fatal — a corrupt manifest has to stay repairable by reinstalling over it. Both paths are covered by red-checked regression tests. * fix(#4249): snapshot Codex skills, agents and VERSION in minimal mode too commit removed from the manifest snapshot. restoreCodexSnapshot is reachable for a core/--minimal install (#2695), and its pass-2 sweeps remove every gsd-* skill dir and gsd-* agent file the snapshot does not claim — so with an empty minimal-mode snapshot a rollback deleted the whole skills/agents surface with nothing to restore it from. Codex resolves skills to $HOME/.agents/skills via the ADR-1239 skills-kind home override, so this is also the reason manifest `skills/` keys do not resolve under configDir: that surface belongs to this snapshot, not to the manifest-driven one. Gate on tomlConfigInstall alone. _codexPreConfigRollback stays null in minimal mode — doing nothing on an early failure is the non-destructive side. Covered by a red-checked regression test that plants bytes in an alternate-home skill file, reinstalls under the core profile marker, and asserts the rollback restores it. * fix(#4249): never remove on rollback unless a prior manifest proves what predates the install Three defects in the manifest-driven Codex rollback, all in its removal half: - ENOENT marked the snapshot usable, arming the removal pass on a FIRST install. GSD may have overwritten a user's file at a manifest-tracked path there, and no prior manifest records the difference — so rollback deleted it where before it merely left it overwritten. Absent, unreadable and malformed manifests now all leave the prior set UNKNOWN and skip removal entirely. - Membership was tested against the map of files whose pre-install read SUCCEEDED, so a tracked file that existed but was unreadable read as introduced-by-this-install and was removed. Track the prior manifest's paths in their own Set and test against that. - The unreachable "delete the manifest when there was no prior one" branch is gone: usable now implies a parsed prior manifest. Also adds the end-to-end test the aggregate gate was missing — the four Codex rollback tests drove the closure directly, proving the restore but not the wiring. installAllRuntimes(['codex','cline']) under an emptied PATH makes Cline's `env node` entry fail validation for real, and asserts Codex's payload comes back. Test preamble (HOME/USERPROFILE sandbox + config-env scrub) is now one helper instead of six copies. Both new tests are red-checked. * test(#4249): use unlinkSync, not rmSync, to drop the manifest in a test lint:ci's raw-fs.rmSync rule points tests at helpers.cleanup for its Windows-EBUSY retry budget. That budget is for directory trees; this removes a single file, which unlinkSync says more precisely and the rule does not flag. * chore(#4249): point the changeset pr field back at the upstream PR * fix(#4249): use an unambiguous dedup key and surface partial-restore failures trek-e's 2026-09-08 adversarial pass flagged two findings in the new entrypoint-validation/rollback code: - assertConfiguredEntrypoints' dedup key already used a raw NUL separator (introduced in ceebb65f2d), but git/Read render NUL as a space, so the key looked like a plain-space join to every reviewer that read the diff. Replace it with JSON.stringify([configPath, scriptPath]) so the separator is visible and unambiguous. - restoreManagedFileSnapshot's per-file restore catch block claimed to 'surface the original error' but only swallowed it, matching (and widening) the pre-existing #3245 restoreCodexSnapshot pattern. Add an actual console.warn using the existing best-effort-warning convention, scoped to just this PR's new function. * fix(#4249): treat a files-less prior manifest as unknown, not known-empty agy's gemini-3.8-flash-high adversarial pass (round 5) found and I reproduced empirically: a structurally-valid manifest missing the files key (e.g. {"version":1}) parses without throwing, so Object.keys(undefined || {}) silently read as 'zero files predate this install' instead of the UNKNOWN state the malformed-manifest guard exists to produce. Rollback's removal pass then deleted every GSD-owned file the failed install's own manifest listed, including ones that predated it — the exact data loss the #4249 CodeRabbit malformed-manifest fix was supposed to prevent, reachable through a JSON.parse success instead of a failure. Route the shapeless case into the same catch-all UNKNOWN path via an explicit shape check. Regression test reproduces the deletion before the fix and confirms the file survives after it. Also extend restoreManagedFileSnapshot's removal-pass rmSync and final manifest-rewrite catches with the same real console.warn trek-e's round-4 review asked for on the per-file restore catch — same rollback function, same operator-facing-signal gap. * docs(#4249): correct which runtimes actually leave a written config on rollback agy's completeness audit (round 5, holistic pass) caught this new paragraph claiming 'for every other runtime, the configuration file(s) already written during that update are left in place' — false for Claude Code and other settings.json-based runtimes, whose write never happens on failure (assertConfiguredEntrypoints runs before finishInstall's writeSettings). Only Cursor/Windsurf/Kimi/Cline actually match that description, since they persist their config file inside install() ahead of the gate. Split the one sentence into the three actual outcomes; matches the PR body's own accurate Before/After wording, which this doc addition had drifted from. * fix(#4249): clean up doc/comment mismatches and dead fields from opus review Opus critical-code-reviewer + ponytail-review pass on the final diff: - assertConfiguredEntrypoints carried finishInstall's old docblock ("Apply statusline config, then print completion message") from before this function was inserted between comment and callee. finishInstall already has its own accurate #4249 comment, so the stale docblock is removed rather than moved. - checked: number on ConfiguredEntrypointValidationResult and error.configuredEntrypointValidation on the thrown error: the first had zero consumers anywhere in the repo, including its own defining file, and is removed. The second matches an existing repo convention (bin/install.js's installerMigrationRollbackFailures, #4249 predates this PR) of attaching structured diagnostic context to a re-thrown Error even before a consumer exists, so it's kept. - finishInstall's own assertConfiguredEntrypoints call is a redundant backstop on the real production path (installAllRuntimes's aggregate call already validates the superset first), but its comment read as though this call alone provided the before-the-write guarantee. Clarified rather than removed — it's the only gate for a caller that invokes finishInstall directly. * chore(#4249): split the manifest-driven rollback engine out into #4544 Issue #4154 asked the installer to consume a validation failure "through the existing rollback mechanism, without a second transaction mechanism". The manifest-driven rollback widening added during review (capture every path the prior gsd-file-manifest.json claims, restore those bytes, remove what only the failed install introduced) is that second mechanism on a plain reading. It is a real fix for a #3245-era gap, but an independent one, so it moves to its own bug report and PR. Removed here: - bin/install.js: the pre-install managed-file capture block and restoreManagedFileSnapshot, plus its call sites in _codexPreConfigRollback and restoreCodexSnapshot (99 lines). - tests/configured-entrypoint-validation.test.cjs: the five tests that exercise the manifest engine. - CONTEXT.md and docs/how-to/update-gsd.md: the sentences describing the widened restore. update-gsd.md again documents the #3245 surfaces only. Kept, because it is #4154's own scope: - the entrypoint-validation gate itself; - Codex's install() result binding rollbackInstallerMigrations to restoreCodexSnapshot (config.toml, hooks.json, skills/gsd-*, agents/gsd-*, gsd-core/VERSION); - the !isMinimalMode gate removal on that snapshot. Binding the closure to the result made it reachable for a core/--minimal install, where its pass-2 sweeps delete every gsd-* skill dir and agent file the snapshot does not claim; an empty minimal-mode snapshot therefore deleted the whole surface with nothing to restore. The surviving aggregate-failure test now asserts on config.toml, a surface the #3245 snapshot owns, instead of gsd-core/CHANGELOG.md, which only the manifest engine restored. Refs #4544 * test(#4249): cover configured entrypoints through the packed install path #4154's scope lists install smoke coverage alongside the installer gate — "assert representative configured entrypoints resolve for supported runtime profiles". The gate itself (assertConfiguredEntrypoints / validateConfiguredEntrypoints) is unit-covered by in-process install() calls; nothing proved the property survives npm pack -> npm install -g -> install.js. Add Cycle 4 to runSmoke. For each of claude and codex — the two distinct config surfaces GSD writes launch paths into (settings.json, and hooks.json + config.toml) — run the tarball-installed installer into a throwaway HOME, then re-read that runtime's own written config and return the new ENTRYPOINT_UNRESOLVED code when a script path it names does not resolve to a file. install-smoke.yml already asserts .code == "ok" on the CLI, so the check becomes a release gate on every matrix host without workflow changes. The scan re-derives paths from the written config instead of reusing the installer's own entrypoint list, and test I shows why that matters: a registration the installer never touched during a run is invisible to the in-process gate, so the install exits 0 and only reading the config back off disk catches the dangling launch path. * ci(#4249): pack a publish-shaped tarball in the install smoke lane `npm pack` runs prepack/prepare (build:lib); only prepublishOnly runs build:hooks. hooks/dist is gitignored, so the tarball install-smoke.yml packs after `npm ci` carries no hook scripts at all — the lane has been smoking a package that differs from the published one in exactly the artifacts the lifecycle smoke is supposed to launch. That went unnoticed because the lane's init runs `--local`, which registers no statusline and therefore registers no hook whose target is missing. A `--global` install on the same tarball exits 1 on #4249's own gate (`gsd-statusline.js (missing)`), which is what the new configured-entrypoint cycle performs, so without this step the cycle would report INIT_FAILED instead of checking anything. Build hooks before packing so the smoked tarball matches prepublishOnly. The CLI now reports 16 configured entrypoints for claude and 1 for codex instead of zero. * fix(#4249): scope Codex's full snapshot restore to entrypoint failures Binding Codex's result to `restoreCodexSnapshot` made ANY finalize-stage exception un-install a Codex install that had already succeeded and already printed its own "Done!" summary — `rollbackFinalizedInstallerMigrations` wraps the whole `finalize()` body, not just the aggregate `assertConfiguredEntrypoints` call. Nothing documents that. `docs/installer-migrations.md#phase-4-installupdate-integration` scopes finalize-stage rollback to installer *migrations* ("the executor uses the journal to restore modified paths"), and this PR's own operator-facing paragraph in `docs/how-to/update-gsd.md` scopes the Codex config.toml/hooks.json/skills/ agents/VERSION revert to entrypoint-validation failures specifically ("If a script is missing, unreadable, ... For Codex, this reverts ..."). The wide behaviour is also incoherent as a transaction abort: the same doc says Cursor, Windsurf, Kimi and Cline keep the config they wrote inside install(). Concretely: `installAllRuntimes(['codex', 'kilo'])` where Kilo's finishInstall hits EACCES writing kilo.json rolled Codex's config.toml back to its pre-install bytes — on an update, silently downgrading a working Codex install to the previous version while the user had just been told it was Done. Select the rollback by error kind instead. `assertConfiguredEntrypoints` already tags its error with `configuredEntrypointValidation`, so the full snapshot restore runs for that error (and anything downstream of it, including finishInstall's per-runtime backstop) and the installer-migrations-only closure runs for everything else. The codex result now also exposes that narrow closure as `rollbackInstallerMigrationsOnly`; `rollbackInstallerMigrations` keeps meaning the full restore, so the direct-call contract asserted by tests/codex-config.test.cjs is unchanged. Adds a regression test that installs codex+kilo together, injects EACCES on the Kilo permission write by monkeypatching node:fs (restored in a finally — never chmod 0o000, which root bypasses in CI), and asserts Codex's config.toml keeps the bytes the successful install wrote. Verified red against the pre-fix unconditional path. Cline cannot host this test: its plan is writesSharedSettings:false + finishPermissionWriter:null, so its finishInstall performs no write and has no non-entrypoint failure path. Kilo's configureKiloPermissions runs unconditionally (unlike OpenCode's, it is not GSD_TEST_MODE-gated) and ends in an unguarded fs.writeFileSync. * docs(#4249): sync CONTEXT.md's rollback description with the round-6 narrowing CONTEXT.md still described Codex's rollback as an unconditional bind to restoreCodexSnapshot after ff13adc00 scoped it to entrypoint- validation failures via rollbackInstallerMigrationsOnly and the configuredEntrypointValidation error tag. Caught during the round-6 PR body pass. * fix(#4249): stop rollbackInstallerMigrations meaning its own opposite Codex's install() result bound `rollbackInstallerMigrations` to restoreCodexSnapshot (the FULL pre-install snapshot restore) and put the actual installer-migrations-only closure behind `rollbackInstallerMigrationsOnly` — so for one runtime the unsuffixed name meant the opposite of what it says, and CONTEXT.md had to concede as much in prose. Invert it: `rollbackInstallerMigrations` is the narrow closure for every runtime, matching both its name and the meaning it already has on next, and the snapshot restore gets its own Codex-only field, `rollbackPreInstallSnapshot`. The selection in rollbackFinalizedInstallerMigrations collapses to one line and no longer needs a fallback chain. Also in this commit, all against the same rollback path: - Correct the rollbackFinalizedInstallerMigrations comment. It read as if the round-6 narrowing prevented any sibling-triggered revert of a Codex install the user has already seen "Done!" for. It does not, and is not meant to: `wide` is true for ANY entrypoint-validation error from ANY runtime, because the aggregate gate is all-or-nothing — an invalid Cline entrypoint reverts Codex's snapshot, which tests/configured-entrypoint-validation.test.cjs's 'an aggregate entrypoint validation failure rolls the Codex install back (#4249)' asserts directly. The discriminator is the error's KIND, not which runtime owns the failing path. Comment and CONTEXT.md now say that. - Name the runtime in the "Configured entrypoint validation failed" error. ConfiguredEntrypointInvalid already carries `runtime`; the message threw it away, leaving an operator of a multi-runtime install unable to tell whose entrypoint broke — which matters precisely because the failure can revert a runtime that was itself fine. - Set `configuredEntrypoints: []` explicitly on the copilot-instructions early return. Every other branch states the key; this one relied on installAllRuntimes' `(result.configuredEntrypoints || [])` defence. `[]` is correct, not a workaround: every Copilot hook is an inline printf one-liner (GSD_COPILOT_*_HOOK_BASH/PWSH), so there is no GSD-managed script or interpreter to resolve. No behaviour change beyond the error-message text. * docs(#4249): narrow the smoke scan's config-surface claim to what it checks RUNTIME_CONFIG_FILES claimed every GSD-managed executable a runtime is told to launch is registered in one of settings.json / hooks.json / config.toml, and that nothing else in a config dir is runtime configuration. Both halves are false as stated. Cline registers its hook at .clinerules/hooks/PreToolUse — a subdirectory, and not one of those names (writeClineArtifacts, src/runtime-hooks-surface.cts). Kimi's native [[hooks]] config.toml lives under resolveKimiHooksTomlDir() (~/.kimi), a directory separate from Kimi's own GSD configDir — the same gap installer-migration 007 already documents as structurally unreachable. The scan is in fact correct for what it runs against: entrypointRuntimes defaults to claude + codex, whose launch paths do all live in those three top-level files. Restate the docstring at that scope, name the two known out-of-scope surfaces, and warn that adding either runtime to entrypointRuntimes without teaching scanConfiguredEntrypoints about its surface yields a scan that finds zero entrypoints and proves nothing. The entrypointRuntimes default comment carried the same overgeneralization ("every other runtime reuses one of them") and is corrected with it. Documentation only; no code change. * fix(#4249): complete configuredEntrypoints/rollback shape on unparseable settings.local.json An internal adversarial review (agy/gemini-3.8-flash-medium, round 8) found that install()'s settings-json early return for an unparseable settings.local.json omitted configuredEntrypoints and rollbackInstallerMigrations from its result, unlike every other branch. rollbackFinalizedInstallerMigrations reads result.rollbackInstallerMigrations unconditionally, so this branch silently dropped its own installer-migration rollback on a later finalize-stage failure. Completed the return shape: configuredEntrypoints: [] (matching Copilot's equally-early no-entrypoints-yet return) and rollbackInstallerMigrations (already in closure scope). Red-then-green regression test added. * test(#4249): ensure hooks/dist before packing in release-tarball-smoke.install.test.cjs Same internal adversarial review (round 8): this suite's before() packed the tarball directly, without the ensureHooksDist() guard every sibling install-test suite (install.test.cjs, install-minimal-hooks.test.cjs, mcp-catalog-parity.install.test.cjs) already uses. On a clean tree, or run in isolation ahead of a suite that builds hooks/dist itself, this suite's pack would ship a tarball with no hook scripts and fail closed on SMOKE.INIT_FAILED instead of testing anything. * fix(#4249): refresh stale test-timings weight for the codex-config split next's own consolidation split (#4139/#4540) moved tests/codex-config.test.cjs's heavy install()-pipeline blocks into tests/codex-config-hooks.test.cjs, but the CI shard packer's weight table (tests/test-timings.json) was never updated: codex-config.test.cjs still carried its pre-split weight (127783ms, ~18x the suite mean), and codex-config-hooks.test.cjs — which now holds the #3245 block this PR extends with its own #4249 install()-pipeline test — had no entry at all, so the packer would silently underestimate it at the table's median weight (roughly a 9x underestimate against its real cost). trek-e's most recent review flagged a Windows shard timeout in-flight on codex-config.test.cjs, plausibly aggravated by this PR's own addition to that file before the rebase moved it. Re-measured both files locally (node --test --test-reporter=tap, max of 3 runs, matching the table's own max-across-streams methodology) and patched just these two entries — not a full regeneration, which would need real multi-lane CI data this session doesn't have access to. * fix(#4249): register configured-entrypoint-validation tests in the conformance-tier lists next's platform-conformance-tier classifier (#4591/#4598) landed after this branch's last rebase, so tests/configured-entrypoint-validation.test.cjs and tests/codex-config-hooks.test.cjs were never classified, failing lint:ci's gen-platform-conformance-tier --check and both the Linux and macOS conformance suites. * fix(#4249): drop codex-config.test.cjs from the #4733 pinned isolated-set expectation next's #4733 (landed after this branch's last rebase) replaced the static ISOLATED_HEAVY_FILES set with a threshold derived live from tests/test-timings.json, and pins the current derived result in EXPECTED_ISOLATED_UNIT_FILES for regression coverage. That pinned list still named codex-config.test.cjs, whose own weight this PR already dropped from 127783ms to 189ms (after splitting its heavy install()-pipeline blocks into codex-config-hooks.test.cjs) — well under #4733's derived 120000ms bar. The live-computed set correctly no longer includes it; the pinned expectation is updated to match. * fix(#4249): name the rollback consequence in the entrypoint-validation error, and prove Cline's file survives it trek-e's review flagged two Major gaps: the thrown error read identically regardless of which of three real outcomes a runtime hit (nothing persisted / snapshot reverted / config left broken on disk), and no test proved the disclosed "left on disk, unreverted" case for Cursor/Windsurf/ Kimi/Cline — only Codex's revert path was ever asserted. assertConfiguredEntrypoints now tags each invalid entry with its actual consequence, mirrored from docs/how-to/update-gsd.md's existing rollback-matrix disclosure. A new test drives the same aggregate failure through Cline (whose own entrypoint is the one that fails) and asserts its hook file is still on disk afterward. * fix(#4249): close 4 gaps antigravity's adversarial review found in the entrypoint-validation PR One review pass (gemini-3.8-flash-high via the antigravity review lane) against this PR's full diff against next, findings independently verified against source before fixing: - Copilot's install() return object was the only one of 6 runtime branches missing rollbackInstallerMigrations — reachable now that this PR's own aggregate gate runs rollback across every result on any runtime's entrypoint failure, not just Copilot's own. - buildHookCommand's unresolved-bash early return skipped track() entirely, so a win32 install with no Git Bash silently produced an unregistered .sh hook instead of the 'unresolved-interpreter' validation failure configuredEntrypointsForHook's own comment said it would. - release-tarball-smoke.cjs reported a Cycle 4 install failure under SMOKE.INIT_FAILED (Cycle 1's code) instead of the already-existing SMOKE.INSTALL_FAILED. - SCRIPT_PATH_RE excluded whitespace to avoid swallowing a shell command's trailing args, which also truncated any configDir containing a space (e.g. a real "/Users/John Doe/.claude"), silently zeroing the scan. Anchored the match on the already-known configDir prefix instead of a generic absolute-path guess: removes the ambiguity outright rather than patching the character class, and stays a raw-text scan on purpose (it catches a writer that emits a path without registering it — a JSON.parse of the expected schema would miss exactly that case). One suggested finding (test-timings.json "missing" the new test file) was verified false — that table only holds measured CI timings, populated after a file's first real run — and one Ponytail suggestion (a JSON.stringify dedup key) was rejected as it would reintroduce a real, if narrow, key-collision risk for no benefit. * fix(#4249): fix fork CI red from a stale changeset pr field and an unquoted docs/ comment changeset-lint requires pr: to match the PR it runs on (16 on the fork, not the eventual upstream number) — rehearsal-branch convention already established earlier in this PR's history. lint-docs-guard-registration's quote-pairing heuristic doesn't require the docs/ path itself to be quoted — it flags a file once ANY quote-delimited span containing "docs/" appears anywhere in it, alongside any real fs read call. A comment ending "...update-gsd.md's rollback-matrix paragraph" supplied the closing quote character (the possessive apostrophe) the heuristic paired with an unrelated single-quoted string earlier in the file. Reworded to avoid the unquoted apostrophe next to the path. * chore(#4249): point the changeset pr field back at the upstream PR Fork rehearsal (PR #16) is green; the real target for this changeset is upstream PR #4249. --------- Co-authored-by: Test <test@test.com> Co-authored-by: Tom Boucher <trekkie@nomorestars.com>	2026-09-16 04:23:53 -04:00
Tom Boucher	740ba0d8a3	fix(#4628): expose DAG-ready plans and restrict dispatch to them (#4781) ... Emitted-Drift-Ack-Growth: execute-phase.md — #4628 consumer wiring: ready_plans parse pointer, not-ready named skip, and waiting condition 2b reference to the ready-wave-gate step file Co-authored-by: sim <sim@local>	2026-09-16 02:38:43 -04:00
Behruz Nassre Esfahani	febe6c9885	fix(#4685): a directory artifact fails its own entry instead of aborting the check (#4735) ... * fix(#4685): a directory artifact fails its own entry instead of aborting the check `must_haves.artifacts` entries are read with `safeReadFile`, which rethrows every errno except ENOENT. A listed path that is a directory therefore threw EISDIR out of the per-artifact loop: `query verify.artifacts` printed Error: EISDIR: illegal operation on a directory, read and reported NOTHING — not the offending entry, and not the plan's other, perfectly checkable artifacts. One directory entry disabled the whole plan's check. Reproduced against a real plan before the fix, and after. A directory is now reported as that entry's own failure, with an issue distinct from `File not found` (the path did resolve; it simply is not the thing an artifact entry can be checked against), and every other artifact in the plan is still checked and reported independently. Anything else the stat or read throws becomes that entry's failure too, carrying its errno, rather than discarding the run — a check that disappears is worse than one that fails, because a failure is visible. Verifying directories properly — matching `contains:`/`min_lines:`/`exports:` across the files inside one — is a feature decision and deliberately not made here, per the issue's stated scope. Authoring-time rejection of a directory path is likewise left alone: the brief raises it as a separate question, and the runtime fix does not depend on it. Also, found in pre-PR review and pre-existing: `safeReadFile(...) || ''` turned a post-stat ENOENT into empty content, so an artifact declaring only `path`/`provides` had no criterion left to fail and passed, having checked nothing. A null read now reports that instead of inheriting a pass. It can only turn a false pass into a failure. Verified: reverting src/verify.cts to the merge-base turns both new rows red with the exact EISDIR message; lint:ci exit 0; full suite 24/24 chunks, 37,280 tests, 0 failures; tests/verify.test.cjs 219/219. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01S4mJZpSNwoyVtRVUQoijfL * chore(#4685): backfill changeset PR number to 4735 Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01S4mJZpSNwoyVtRVUQoijfL * test(#4685): pin the injected-I/O branches, and narrow the guard's comment Review findings from #4735. Major — the two error branches this PR adds were untested, and the PR body claimed the mid-check ENOENT window was "not deterministically reproducible through the CLI seam these tests drive." That was wrong: ADR-3574 records this repo's convention for exactly this — inject filesystem failures by monkeypatching the fs method, never by chmod or mode-bit tricks, which root bypasses and yields a test that passes with zero coverage in root Docker and CI. Both branches are now pinned that way. The injection runs in the CHILD via NODE_OPTIONS=--require, because `output()` writes fd 1 directly (`writeAllSync(1, …)`, io.cjs) rather than through console.log, so an in-process call cannot have its JSON captured. The preload patches the child's own module objects, which the compiled code reads at call time. - a file that disappears between stat and read now fails as that entry rather than passing on empty content - a non-ENOENT errno (EACCES) is reported as that entry's failure, carrying its code, so an operator can tell a permissions problem from an I/O one The injection matches the target by path SUFFIX, not string equality: the first cut compared absolute paths, and a /tmp vs /private/tmp prefix difference silently disarmed it — the test passed while asserting nothing. A disarmed injection test is worse than no test, so the reason is recorded at the call site. Nit — the comment above the try block said the guard "covers what the stat and read below actually throw", which reads as if the min_lines/contains/exports checks inside the same block were deliberately guarded too. They are pure string operations and cannot throw; the comment now says so rather than implying a guarantee it does not make. Verified: reverting src/verify.cts to the merge-base turns all three #4685 rows red — the directory row and both new ones; lint:ci exit 0; full suite 24/24 chunks, 37,389 tests, 0 failures; tests/verify.test.cjs 221/221. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01S4mJZpSNwoyVtRVUQoijfL --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com> Co-authored-by: Tom Boucher <trekkie@nomorestars.com>	2026-09-16 01:39:52 -04:00
Twisted Fate	ecc508139a	fix(#4730): decode entity-escaped ampersands before verify-command-paths segment splitting (#4755) ... * fix(#4730): decode entity-escaped ampersands before verify-command-paths segment splitting Planners emit <automated> bodies with the chain operator entity-escaped (`&amp;&amp;`), and the executing agent reads the decoded (rendered) form. The grounding probe split the raw text on &&/||/;/newline without decoding first, so `&amp;&amp;` was cut at its semicolons and a cd-form target absorbed the trailing `&amp` fragment — an existing directory was reported missing_dir (blocker), feeding false blockers into the revision loop. Decode `&amp;` → `&` inside resolveVerifyCommandTarget, after result.command captures the text verbatim and before any segment splitting or target resolution, so the escaped and literal forms of the same command produce identical verdicts. Module-private helper beside splitSegments, mirroring the sibling src/verify.cts decodeEntityAmps (#3611); the two gates are separate modules and neither imports the other. Regression coverage pins escaped/literal verdict parity for: existing dir + manifest (ok), missing dir (missing_dir blocker kept), dir without manifest (no_manifest blocker kept), --prefix form, a literal & inside a quoted dir name, and a full probePhaseVerifyCommands pass whose reported command field stays verbatim. * docs(#4730): use the documented pr:0 placeholder in the changeset fragment The fragment carried pr: 4730 — the ISSUE number, the exact guess-shape DEFECT.CHANGESET-PR-FIELD-DRIFT (#3316, #3325) exists to catch: it parses as a positive integer so local lint passed, but on a real PR run the drift check would fail it against the actual PR number. CONTRIBUTING.md documents pr: 0 as the deliberate unresolved placeholder used during initial commit before the PR number exists (scripts/changeset/new.cjs accepts 0 for exactly this reason); the gate's fail_invalid_fragment on an unbackfilled 0 is the designed backfill enforcement, not a defect. No production or test changes. Backfill pr: with the real PR number once the PR is created. * docs(#4730): backfill changeset pr field with the real PR number pr: 0 → pr: 4755 (the PR carrying this fix), completing the documented placeholder workflow; the changeset gate's content validation can now pass. --------- Co-authored-by: TwistedRiCen <16397953+TwistedRiCen@users.noreply.github.com> Co-authored-by: Tom Boucher <trekkie@nomorestars.com>	2026-09-16 00:36:10 -04:00
0xdhx	092d9256b8	fix(#4748): carry a letter-suffixed phase id through the seven shell sites that aborted or truncated it (#4768) ... * test(#4748): pin the letter-axis defect at the seven shell sites outside #4660's six Extends tests/nsegment-phase-grammar.test.cjs one class over: for each of the seven sites the live shell lines are read off disk by anchor and executed in bash against a letter-suffixed fixture. The four `$((10#$PHASE_INT))` split sites must yield PHASE_N without a shell error for `03A` / `12A` / `3A` / `03A.1.2` and the commit-scope ERE they build must match both `feat(3A-01):` and `feat(03A-1):`; the review-file lookup must bind init's `padded_phase` rather than re-pad in shell; the `--from`/`--to`/`--only` and plan-review-convergence extractions must return `12A` / `23A.1.2` (and `23.1.2`) whole; the legacy normalizer must pad `3A` to `03A` and must not mangle an already-padded `08`. Every pre-existing shape (`06`, `08.5`, `23.1.2`, `36.14`) is a regression control. tests/init.test.cjs asserts `init execute-phase` emits `padded_phase` for a directory-backed `03A`, a ROADMAP-only `4B` (→ `04B`), the existing ROADMAP fallback `1` (→ `01`), and `null` when the phase is not found. Negative control against the unfixed tree: 41 failures in the grammar file, exactly the "(fails before the fix)" cases and the three derived from them (scope ERE, three-flag extraction, the `08` octal trap); 2 in init.test.cjs, both the new assertions. Every regression control already green. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_017cjzdZtYjcBAa3Lqh2VrLK * fix(#4748): carry a letter-suffixed phase id through the seven shell sites that aborted or truncated it The canonical phase-number grammar (src/phase-id.cts) is digits, an optional uppercase letter, then dotted segments — `12A`, `3A`, `23A.1.2` are documented shapes that `init`, `phase-id.cts` and `phase remove` renumbering already round-trip. Seven shell sites in shipped workflows and references still assumed digits-and-dots. Four classes, one fix each: Class 1 — `PHASE_INT=${PHASE_NUMBER%%.*}; $((10#$PHASE_INT))` (execute-phase.md ×2, completion-reconciliation.md, tdd.md). The post-#4619 split stops at the first DOT, so on `03A` the "integer" is `03A` and bash aborts with `value too great for base`. Split at the first NON-DIGIT instead (`%%[!0-9]*`): the integer half is a pure digit run, and the letter rides along in the rest the way the dotted fraction already did — `03A.1.2` → PHASE_N `3A\.1\.2`, so the #4003 zero-pad-tolerant scope ERE matches both `feat(3A-01):` and `feat(03A-1):`. Byte-identical output for every id that worked before. Class 2 — `PADDED=$(printf "%02d" "${PHASE_NUMBER}")` before the REVIEW.md lookup (execute-phase.md). `printf` cannot pad a letter id (prints `03`, exits 1) — and cannot even re-pad an already-padded `08`, which bash reads as an invalid octal and prints as `00`, so the lookup resolved phases 08 and 09 to `00-REVIEW.md` today. The disk path hands the workflow the directory's padded number but the ROADMAP fallback hands it the heading's bare one, which is why the re-pad existed. `cmdInitExecutePhase` now emits `padded_phase` through `normalizePhaseName`, exactly as the plan-phase and code-review inits do, and the workflow binds `{padded_phase}` instead of re-deriving. Class 3 — `grep -oE '[0-9]+\.?[0-9]*'` (autonomous.md `--from`/`--to`/`--only`, plan-review-convergence.md). Stops at the letter, so `--from 12A` ran from phase 12 with no error. Now the canonical ERE `[0-9]+[A-Z]?(\.[0-9]+)*`, which also closes the single-segment dot-axis gap the same shape carried (`23.1.2` → `23.1`, #4568's class in a spelling neither lint saw). Class 4 — the legacy manual normalizer (phase-argument-parsing.md, reached from mvp-phase.md). Its two branches (`^[0-9]+$`, `^[0-9]+\.[0-9]+$`) left `12A` unpadded and never padded `3A` to the `03A` a directory carries; its integer branch also hit the same `printf` octal trap on `08`. One branch for the whole canonical token now, padding the digit run via `$((10#…))`. Whether this legacy surface should instead be retired in favour of `init`'s normalization is the maintainer call the issue names; extending it keeps the documented contract true either way. Driven end to end: `init execute-phase 3A` on a fixture with a `03A-letter-variant/` directory emits `phase_number: "03A"` and now `padded_phase: "03A"`; on a ROADMAP-only `### Phase 4B:` it emits `"4B"` / `"04B"`. The issue's own evidence line claimed `padded_phase` was already in the execute-phase init output — it was not; that key is emitted by the code-review / plan-phase inits, which is where the claim was read from. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_017cjzdZtYjcBAa3Lqh2VrLK * chore(#4634): extend lint-phase-id-drift with three ratchets for letter-hostile phase-id consumers The rules that landed with #4619, #4568 and #4660 police grammar MIRRORS — regexes that describe a phase id. The #4748 sites are CONSUMERS of one, and every existing rule reported clean on them: the shell-arithmetic rule's `_INT` escape trusts a NAME the dot-only split did not earn on `03A`; the `[0-9]+\.?[0-9]*` shape is neither the bounded form the single-segment rule bans nor the unbounded form the letterless rule inspects; and nothing looked at `printf "%02d"` at all. Three narrow additions, one per shape: - findDotOnlyIntegerSplitDrift — `X_INT=${<phase-var>%%.*}`; the safe split is `%%[!0-9]*`. Keys on the SOURCE variable being phase-carrying. - findLooseDottedPhaseRegexDrift — `[0-9]+\.?[0-9]*` / `\d+\.?\d*` on a phase-carrying line; the canonical form is `[0-9]+[A-Z]?(\.[0-9]+)*`. Disjoint from the two sibling regex rules by construction. - findShellPhasePrintfPadDrift — `printf "%0Nd" …` whose arguments name a phase-carrying, non-`_INT` variable; a pad of an `_INT` via `$((10#…))` and a `{padded_phase}` binding are the sanctioned shapes. Same `<!-- phase-id-owner: … -->` sanction, same scan roots as their nearest sibling (shell idioms over workflows + references, the regex shape over workflows + references + agents), same documented limit of a per-line textual scan. The post-#4619 comment that described the `_INT` convention as proven by `%%.*` is corrected to name the digit-run split. Confirmed against the base commit: each rule fires on exactly its own unfixed sites (2+1+1, 3+1, 1+1) and zero violations remain on the fixed tree. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_017cjzdZtYjcBAa3Lqh2VrLK * docs(#4748): add Fixed changeset Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_017cjzdZtYjcBAa3Lqh2VrLK * chore(#4748): refresh the compact-content benchmark baseline and acknowledge emitted growth The three top-level workflow files below grew by the letter-aware split, the canonical extraction ERE, the `{padded_phase}` binding, and the comment lines that name the grammar each site now honours. The committed compact-content benchmark moved with them; refreshed with `benchmark-compact-content.cjs --write` (aggregate reduction 15.47% -> 15.45%). Emitted-Drift-Ack-Growth: execute-phase.md — #4748: first-non-digit PHASE_INT split at the plan-selection and TDD-gate sites, `{padded_phase}` binding at the REVIEW.md lookup, and the comments naming why (482 bytes) Emitted-Drift-Ack-Growth: autonomous.md — #4748: canonical `[0-9]+[A-Z]?(\.[0-9]+)*` at the --from/--to/--only extractions plus one comment naming the grammar (249 bytes) Emitted-Drift-Ack-Growth: plan-review-convergence.md — #4748: canonical `[0-9]+[A-Z]?(\.[0-9]+)*` at the phase extraction plus one comment naming the grammar (160 bytes) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_017cjzdZtYjcBAa3Lqh2VrLK * fix(#4748): name padded_phase in execute-phase.md's init parse list A `{field}` token inside a workflow bash block is substituted from the init JSON only for fields the workflow tells the model to parse. `phase_number` is on that list; `padded_phase` was not, so the `PADDED="{padded_phase}"` binding at the review lookup would have been a literal — for every phase, not only letter ones. Found by the pre-file adversarial review (claim 2, the author's own named suspicion); the test now asserts the parse list carries the field beside `phase_number`. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_017cjzdZtYjcBAa3Lqh2VrLK * chore(#4634): key the dot-only split rule on its source and widen the printf rule to any %d form Two false negatives from the pre-file adversarial review of the three #4748 ratchets: `PHASE_PREFIX=${PHASE_NUMBER%%.*}` escaped the split rule because the destination did not end in `_INT` (the defect is the split, not the name it lands in), and `printf '%02d'` / `printf "%2d"` escaped the printf rule because it required double quotes and the zero flag (`%d` cannot parse a letter id under any width). Both rules now key on the phase-carrying SOURCE alone; base-site firing counts are unchanged (2+1+1, 1+1) and the fixed tree stays at zero. The `[[:digit:]]` spelling and the `/phase/i` heuristic remain the sibling rules' documented limits. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_017cjzdZtYjcBAa3Lqh2VrLK * chore(#4748): refresh the compact-content benchmark baseline after the parse-list edit Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_017cjzdZtYjcBAa3Lqh2VrLK * test(#4748): compose init's emitted padded_phase through the live REVIEW.md lookup The Class 2 site is a `{padded_phase}` template token, which no test can execute as written. This substitutes the value init emits (`normalizePhaseName`) into the three live lookup lines and runs them against a fixture, so the emitted value, the binding, the path construction and the status extraction are exercised together — `03A-REVIEW.md` and `08-REVIEW.md` each resolve to their own status. Suggested by the resumed adversarial review pass (claim C). Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_017cjzdZtYjcBAa3Lqh2VrLK * test(#4748): move the #4619 and #4003 source-parity pins to the letter-safe split tests/execute-phase-decimal-arithmetic.test.cjs and tests/safe-resume-gate-anchoring.test.cjs pin the four Class 1 sites' snippet byte-for-byte, so the first-non-digit split reddened both in the whole-suite run (scripts/ci-test-scope.cjs does not select either file for a workflow edit — the scoped run was green). The pinned snippet is now the shipped one, and the behavioural half of the #4619 file gains the letter case (`03A` → `3A`, `23A.1.2` → `23A\.1\.2`) beside its decimal cases. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_017cjzdZtYjcBAa3Lqh2VrLK * chore(#4634): key the dot-only split rule on the _INT destination again, tolerating the quoted spelling Keying on the source alone (the previous commit's widening, from a review probe) flags `PARENT_PHASE="${PHASE_NUMBER%%.*}"` in gap-closure-artifacts.md — a correct derivation that wants everything before the first dot, letter included. The defect this rule polices is a dot split INTO the name the shell-arithmetic rule trusts as an integer, so `_INT` is the discriminator on purpose; the quoted spelling that site uses is now tolerated so the same shape into an `_INT` cannot hide behind it. Base-site firing unchanged (2+1+1), zero on the fixed tree, and the parent-phase line is pinned as a silent case. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_017cjzdZtYjcBAa3Lqh2VrLK * test(#4748): use t.after() for the composition test's fixture cleanup CONTRIBUTING forbids try/finally inside a test body; the per-test cleanup form is `t.after(() => cleanup(dir))`. Flagged by the filing driver's test-ruleset gate before the PR was created. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_017cjzdZtYjcBAa3Lqh2VrLK * chore(#4748): set changeset fragment pr to 4768 * chore(#4748): refresh the compact-content benchmark baseline after rebasing onto next Regenerated with `node scripts/benchmark-compact-content.cjs --write` on the rebased tree (base 0d6bf19bf); `--check` confirms it matches the live recompute. Only the execute-phase split and the aggregate totals differ from next's copy. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_017cjzdZtYjcBAa3Lqh2VrLK --------- Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com> Co-authored-by: Tom Boucher <trekkie@nomorestars.com>	2026-09-15 23:31:36 -04:00
0xdhx	25d1cb916f	fix(#4721): give worktree cleanup-wave's merge its own timeout, report merge_timed_out, and restore the index a killed merge leaves staged (#4766) ... * fix(#4721): give cleanup-wave's merge its own timeout, report merge_timed_out, and restore the index a killed merge leaves staged `worktree cleanup-wave` ran `git merge --no-ff` under the module-wide DEFAULT_GIT_TIMEOUT_MS (10 s) that is sized for plumbing calls. The merge is the one call in the wave that runs user hooks, so a repo whose pre-merge-commit hook is a test-suite gate lost every code-bearing executor merge. Three things went wrong at once, each fixed here: 1. Budget. The merge now passes an explicit timeout — DEFAULT_MERGE_TIMEOUT_MS (10 min), overridable via deps.mergeTimeoutMs. Every other git call in the wave keeps the module default; the shared constant is untouched, because every other caller is exactly what its 10 s comment describes. 2. Reason. A merge that does time out blocks on `merge_timed_out`, and its stderr names the budget and says the hook may still be running, instead of `merge_failed` carrying whatever the hook had printed before git was killed — which made a healthy executor branch look broken. 3. Residue. A merge killed during its hook has already staged the merged tree into the primary's index but never wrote MERGE_HEAD, so `git merge --abort` finds nothing and repoRootStillMidMerge (#2852) reads the primary as clean while the executor's whole diff sits staged against the old HEAD; a `git commit` from that state squashes the executor's history into one parent. After any failed merge the wave now reads `git diff --cached --name-only`; anything staged is the merge's own (git refuses to start a merge when the index differs from HEAD), so it runs `git reset --merge` — restores exactly those paths, keeps unrelated unstaged edits — and re-reads. Restored paths are reported as WAVE_CLEANUP_WARNING.MERGE_RESIDUE_RESTORED and the wave continues; a still-dirty or unreadable index reports MERGE_RESIDUE_LEFT_STAGED and halts the remaining entries, the same repo-level carve-out an unfinished merge takes. Tests: five mock-driven rows (budget wiring incl. the deps override, the timeout classification with restore, the no-reset control for an ordinary refused merge, an unrestorable residue halting the wave, an unverifiable index failing closed) plus a real-git row that runs a sleeping pre-merge-commit hook under a 1 s budget and asserts HEAD unmoved, index and worktree clean, the executor branch intact — with the same fixture merging cleanly under the default budget as its negative control. Two existing #2852 rows gain a handler for the new post-failure index read. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01TbbqrGJMuiuLftAMVLayb9 * docs(#4721): add Fixed changeset Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01TbbqrGJMuiuLftAMVLayb9 * test(#4721): release the real-git fixtures with t.after, not try/finally The two real-git rows cleaned up their scratch repo in a `finally` block; this file's own convention for fixture teardown is the test context's `t.after(() => cleanup(dir))`, and the house PR ruleset flags `finally` in a test body. Behaviour-neutral. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01TbbqrGJMuiuLftAMVLayb9 * fix(#4721): gate the residue restore on the timeout, re-apply a merge autostash, and correct the hook census Three findings from the pre-file adversarial review of the previous commit, each driven on real git before changing code: 1. A merge git REFUSED ("your local changes … would be overwritten") also leaves no MERGE_HEAD — and that refusal is exactly what a pre-existing dirty primary index earns. The residue restore read that index as the merge's own and `reset --merge`d the operator's staged work away (driven: a staged edit to an unrelated file was discarded and reported as "restored"). The restore now runs ONLY when the merge timed out; a refusal is an immediate exit, never a timeout, so on that path nothing is read or reset. 2. `merge.autoStash=true` lets a merge start on a dirty index by parking the work in MERGE_AUTOSTASH, which a killed merge never re-applies. `git reset --merge` moves that stash into the stash list; the wave now runs `git stash pop --index` afterwards (the outcome `merge --abort` gives an autostashed merge), and reports WAVE_CLEANUP_WARNING.MERGE_AUTOSTASH_UNRESTORED (path null) when the pop fails or the autostash state could not be read — the work stays in the stash, the index is clean, the wave continues. Because of this the reset runs on a timed-out merge even when the index reads clean. 3. The merge is not the only hook-running git call in the module: `worktree add` runs post-checkout and every ref update runs reference-transaction. It is the only call that runs the commit-family hooks, which is what the budget is for. Comments and docs say so now. Tests: the "ordinary merge_failed" control becomes the regression row for finding 1 (strict mock — a `diff --cached` or `reset --merge` on a refused merge throws), plus a mock row for the autostash pop (dirty and clean index, pop success and failure), and two real-git rows: a refused merge over pre-existing staged work leaves it byte-identical, and a killed merge under merge.autoStash restores the executor residue AND puts the operator's staged work back. The real-git hook now sleeps 4 s against a 1.5 s budget for margin on slow runners. The two #2852 handlers added earlier are removed — the residue read no longer fires on their path. Negative control: 4 of the 10 #4721 rows fail on the previous commit. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01TbbqrGJMuiuLftAMVLayb9 * fix(#4721): key the residue restore on a killed merge, and re-read the index after a failed autostash pop Two more findings from the continuation review, both driven: 1. An externally delivered SIGTERM leaves the same staged/no-MERGE_HEAD state as the timeout, and the seam reports it as exitCode null + signal with timedOut false — so the timeout-only gate skipped the restore on a state it was written for. The gate is now "killed": timedOut, or a null exit code with a signal. A refused merge still exits with a code and is still never touched. The reason stays merge_failed for a signal kill. 2. A failed `git stash pop --index` keeps the stash entry but can leave conflict entries (UU) and partially applied paths, after which the next merge fails on "you have unmerged files"; the code returned halt:false on the strength of the pre-pop recheck. The index is now re-read after a failed pop and a dirty result halts the wave as merge_residue_left_staged alongside the merge_autostash_unrestored warning. Also driven and now documented rather than changed: a kill that lands once MERGE_HEAD exists (inside commit-msg) is the ordinary #2852 abort path — `git merge --abort` restores the tree and re-applies an autostash itself, unstaged, as git does for any aborted autostashed merge. Tests: the pop-failure mock row now asserts the post-pop re-read and gains a conflict-leftover variant that halts; a signal-kill mock row; a real-git row with the sleeping hook moved to commit-msg (timed out, no residue warnings, MERGE_HEAD cleared, primary clean). 414 pass. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01TbbqrGJMuiuLftAMVLayb9 * fix(#4721): key the kill gate on the seam's signal, not on a null exit code The shell projection seam normalizes a signal death to exitCode 1 and carries the signal alongside (`_spawnResult`: `result.status ?? 1`), so the previous `exitCode === null && signal` gate could never fire in production and the unit row that covered it modelled a shape the seam does not emit (caught in the round-3 review). The gate is now `timedOut || signal`; a refused merge exits with a code and no signal. The mock row uses the real shape, and a mocked spawnSync signal death driven through the compiled seam reaches `reset --merge` and reports the residue restored. Also: three comments that still said "at its budget" / "runs user hooks" / "the index is clean", and the CLI-TOOLS sentence that reserved `merge_failed` for refusals and conflicts, now name the signal case. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01TbbqrGJMuiuLftAMVLayb9 * chore(#4721): set changeset fragment pr to 4766 --------- Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com> Co-authored-by: Tom Boucher <trekkie@nomorestars.com>	2026-09-15 22:28:09 -04:00
Tom Boucher	b5fc8061b3	fix(#4624): persist orchestrator-worktree worker lifecycle records (#4778) ... * fix(#4624): persist orchestrator-worktree worker lifecycle records * fix(#4624): address review findings on the worker lifecycle protocol * fix(#4624): require the summary path and surface torn records on status --path * fix(#4624): distinguish no-record from torn-record, tolerate older shims in the sweep * docs(#4624): backfill changeset PR number --------- Co-authored-by: sim <sim@local>	2026-09-15 20:25:24 -04:00
Tom Boucher	779f67cb11	fix(#4378): mint collision-free SEED-YYMMDD-xxx seed ids instead of a shared count (#4754) ... * test(#4378): regression tests for collision-free seed ids * fix(#4378): mint collision-free SEED-YYMMDD-xxx ids, not a shared count plant-seed derived the next seed id from 'ls .planning/seeds/SEED-*.md | wc -l'. .planning/seeds/ is shared but each worktree only sees what has merged, so two workstreams planting before either merges computed the same id and git merged both files silently. The id is now the local date plus a 3-char random base36 suffix -- the shape .planning/quick/ already uses -- computed from knowledge one worktree has alone, with a same-day regen guard. deriveSeedIdentity learns the new canonical grammar alongside legacy SEED-NNN (whose parsing never changes), the --enrich parser and the filename-prefix fallback keep the full new-format id, and the docs that state the filename shape move to it. The prefix fallback previously truncated any non-pure-numeric id at 'SEED-<digits>' -- the same one-id-two-answers ambiguity the issue reports, reproduced one level down. * fix(#4378): harden seed id generation per adversarial review - parse-idea: anchor the --enrich extractor to the flag and capture the complete id, uppercase-tolerant; a leftmost 'SEED-[0-9]+' truncated an uppercase or malformed suffix to its date and enriched an arbitrary same-day seed via head -1. Ambiguous and unmatched targets now fail closed instead. - generate-seed-id: tolerate the expected SIGPIPE under pipefail, abort loudly when the suffix cannot be drawn (an empty suffix would collapse every seed's id to the bare date), and run the same-day regen guard as a find existence test (the 'ls <glob>' shape trips the #3409 drift guard and degenerates under a stray nullglob). - deriveSeedIdentity: document the theoretical legacy/new grammar ambiguity (6-digit counter + 3-char base36 slug, no frontmatter). - changeset: state the residual same-day collision bound instead of implying zero. Emitted-Drift-Ack-Growth: plant-seed.md — the counting step became hardened date+random generation with explicit failure modes; growth is the failure handling, not duplicated logic * fix(#4378): address standards and spec review findings - tests: move the allow-test-rule marker to its suppression site (the file-header placement was inert per CONTRIBUTING site-scoping); add width-boundary coverage (5/7-digit dates, 2/4-char suffixes pin the documented branch behavior); add a writer-to-reader parity property that parses the mint widths out of the shipped workflow so the two grammar owners cannot drift; cover uppercase ids end-to-end in the reader. - plant-seed.md: draw/retry restructured as one loop with a loud terminal failure; SEED_SUFX renamed SEED_SUFFIX; regen guard drops the redundant head -1; the ambiguity error no longer advises an impossible 'complete id' for duplicate legacy ids. - commands.cts: refresh the cmdListSeeds comment still describing SEED-NNN as the only canonical form. - changeset: drop the audit claim the spec axis showed to be an overstatement (audit's id display is filename-derived, pre-existing). - remove a stray untracked artifact file swept into the tree. * test(#4378): correct boundary expectations to the module's real branch behavior The first matrix run on the boundary tests caught my hand-trace of the regex branches, not a module defect: the slug regex's alternation backtracks to the legacy branch whenever the canonical branch cannot complete (so the slug is the remainder after the legacy numeric prefix), and the 7-digit case fails the canonical branch at its 7th digit before the dash. Pin the verified values. * docs(#4378): backfill changeset PR number * fix(#4378): audit seed identity uses the canonical grammar Review of this PR found the audit surface publishing a fused filename stem (SEED-081-region for SEED-081-region.md) where list-seeds reports the canonical id -- one id, two answers across surfaces, the same ambiguity class the issue files. scanSeeds now derives identity through the SAME deriveSeedIdentity the list-seeds gate uses (frontmatter id, then filename id-prefix, then stem), and audit-open acknowledge resolves --seed-id by scanning for the derived identity, falling back to the literal stem so callers scripted against pre-canonical output keep working. Roll-in per the fix-inline rule: found during this PR's review, same seed-identity seam. RED probe: pre-fix audit published seed_id SEED-081-region-becomes / slug 081-region-becomes for a legacy seeded file; post-fix SEED-081 / region-becomes, matching list-seeds. * test(#4378): probe timeout uses the class norm after windows-lane timeout The windows conformance shard failed its bounded sh -c probes at the local 5000ms bound (cold sh.exe spawn under shard load) while the identical code passed this PR's two earlier windows waves. The probe now uses PROBE_TIMEOUT_MS from the class-norm module instead of a local override, per the helpers/timeouts.cjs convention. --------- Co-authored-by: sim <sim@local>	2026-09-15 11:41:29 -04:00
Tom Boucher	cbbde6786a	fix(#4546): deferred UAT follow-ups no longer block completion and promote to the backlog (#4769) ... * test(#4546): failing-first tests for deferred uat follow-ups * chore(#4546): regenerate derived lists for the deferred-promotion suite The new verify-work-deferred-promotion suite changes the tests/ tree the macOS conformance-tier classifier tracks and is a novel file under the verify prefix in the test-file-count ratchet; both derived lists are regenerated/registered per their own guards' instructions. * fix(#4546): deferred uat follow-ups no longer block, and get promoted Two halves of one disconnect (#1921's deferral design vs the completion predicate): - uat-predicate: the item parser now captures the block's reason: line alongside result:. A skipped item whose reason carries the verify-work writer's 'Deferred follow-up:' template is a deliberate deferral -- non-blocking, flagged deferred in the report. Quote- tolerant (the writer wraps the value) and case-insensitive. A reasonless skip, a non-deferral reason, pending/blocked/issue/ failed/missing all still block, exactly as before. - verify-work complete_session: when the Deferred Follow-Ups section is non-empty, offer to promote the items to a ROADMAP.md 999.x backlog entry reusing next.md's prior_phase_completeness entry shape, with a --files-scoped commit. Offer, not auto-mutation -- matches the workflow's interactive convention and next.md's own prompt style. * chore(#4546): refresh compact-content benchmark baseline verify-work.md grew (the #4546 deferred-follow-up promotion offer in complete_session); the registered split's token counts moved with it. Baseline recomputed with the script's own --write. Emitted-Drift-Ack-Growth: verify-work.md — complete_session gained the deferred-follow-up promotion offer (detection, [P]/[K] choice, the next.md-shaped 999.x entry template, and the --files-scoped ROADMAP.md commit); the growth is the new contract text, not duplication * fix(#4546): gate/audit agreement and review fixes for deferred follow-ups - src/uat.cts categorizeItem: a skipped item carrying the deferred follow-up template reason now categorizes as 'deferred' (the category already existed for deferred-items.md entries) instead of being misfiled into the blocked families by keyword match -- the gate/audit agreement #3078-CR expects, restored in the permissive direction the #1921 design intends. Checked BEFORE the keyword families so '... on the release build next version' is not build_needed. - verify-work.md promotion step: numbering scans for the smallest free 999.n (count races + non-contiguous history), one backlog entry per deferred follow-up, ROADMAP.md-absent behavior specified, idea text newline-flattened, Deferred at placeholder harmonized with next.md. - DEFERRED_REASON_RE: trust assumption documented (authoring contract, not a security boundary; non-matching spellings block fail-closed). - tests: the property now drives evaluateUatPassed and derives expectations from the input spec (never restates the matcher), includes the no-result-line branch, and pins its seed; the parity test drops try/finally for the approved pattern, uses createTempDir, sites its allow-test-rule marker at the suppression site, and asserts the literal [P]/[K] choices. * fix(#4546): close promotion-test docstring, drop fc replay-path misuse, refresh baseline The final matrix run caught three defects in my own review-fix commit: the parity test file's JSDoc was left unterminated (the whole file parsed as one comment -- zero tests registered, hence the file-level 'test failed' the runner reported); fast-check's replay-path parameter was misused as a label (invalid path at replay); and the workflow-text ambiguity fixes re-drifted the compact-content benchmark baseline. * docs(#4546): add Fixed changeset for deferred follow-up coverage * docs(#4546): backfill changeset PR number * fix(#4546): use the pattern seam escapeRegex for shape-marker matching The hand-rolled metacharacter escape in the shape-marker assertion tripped local/no-adhoc-regex-escape, whose named remedy this adopts. --------- Co-authored-by: sim <sim@local>	2026-09-15 07:30:38 -04:00