12 Commits

Author	SHA1	Message	Date
Tom Boucher	dd4f179672	feat(#3970): per-task external-tracker content-resolution seam (#4000) ... * feat(#3970): per-task external-tracker content-resolution seam Implements ADR-3646 (Phase 1, #3970): a `<task tracker-id="...">` attribute plus a new optional `taskContentResolver` capability-manifest field let a capability resolve a task's action/verify/acceptance-criteria/read_first/done content from an external issue tracker instead of PLAN.md's inline body. - src/plan-document.cts: parses the `tracker-id` attribute into `PlanTask.trackerId` - src/task-content-resolution.cts: new leaf module — split/find/build/resolve, with a hard-halt (throw) contract on ambiguous/failed/timeout/malformed resolution, never a silent fallback to possibly-stale inline text - src/task-command-router.cts: new `task resolve-content --plan --task-id --raw` CLI verb wiring the module into a real process exit code - gsd-core/bin/lib/capability-validator.cjs: validates the new `taskContentResolver` manifest field (feature-role only, cross-capability trackerPrefix uniqueness) - gsd-core/workflows/execute-plan.md, gsd-core/references/loop-hook-dispatch.md, docs/reference/capability-manifest.md: wire the seam into the per-task loop and document it as a new `execute:task` point outside the existing contribution/step/gate vocabulary (unconditional in autonomous mode) Closes #3970 * fix(#3970): gate checkpoint tasks out of content resolution, close trackerPrefix grammar parity gap, cover path-traversal guard Standards/Spec code-review pass on the task-content-resolution seam (ADR-3646 Phase 1) found three defects: 1. execute-plan.md's task-content-resolution bullet fired on any tracker-id-bearing task with no check that it wasn't type="checkpoint:*", contradicting ADR-3646 Decision 1 (a checkpoint task must never enter resolve-content). plan-document.cts already parses trackerId: null unconditionally for checkpoint tasks; only the workflow prose needed the fix, so the bullet now explicitly excludes checkpoint tasks. 2. task-content-resolution.cts's parseResolverDeclaration accepted any non-empty trackerPrefix with no grammar check, while capability- validator.cjs's KEBAB_RE enforces kebab-case at install time — a Generative Fix Divergence gap. Added the same grammar (as a literal regex, documented as intentionally not shared across the .cts/.cjs build boundary) plus a parity test asserting the two surfaces agree across a valid/invalid trackerPrefix table. 3. task-command-router.cts's routeResolveContent path-traversal guard on --plan had zero test coverage. Added a test exercising a ../../../etc/passwit-shaped path and asserting the USAGE rejection names the offending path. * fix(#3970): sanitize resolver diagnostics and cap resolver timeoutMs Two findings caught by an isolated security-review pass on the task content resolution seam: - ResolverFailedError/ResolverMalformedOutputError embedded raw, unsanitized subprocess stderr/stdout (attacker/model-influenced via the tracker-id argv token) into .message. A hostile or buggy resolver could smuggle a newline plus a forged "Error: " line, or terminal escape sequences, into a diagnostic io.cjs's error() writes verbatim to stderr. Fixed at the constructor (task-content-resolution.cts) via io.cjs's existing formatDiagnosticToken(), so every caller of resolveTaskContent gets a safe .message by construction. - capability-validator.cjs's validateTaskContentResolverFields had no upper bound on taskContentResolver.invoke.timeoutMs, letting a manifest declare an effectively unbounded value and defeat the "bounded subprocess" design intent. Added a 120000ms ceiling specific to this field, without touching the shared isPositiveIntegerMs() helper (still used unbounded by the reviewer lane's timeoutFloorMs and probe timeoutMs). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> * fix(#3970): fix gsd-test failures — stale prose allowlist line and stderr-vs-message assertion gsd-test (remote dockerized matrix) came back red with 5 failures on this PR; all five are real defects, fixed here. - tests/no-bare-gsd-tools-command-position.test.cjs: PROSE_ALLOWLIST's execute-plan.md entry pointed at line 415, which ffc190df4's checkpoint-exclusion caveat (added near line 221) shifted down by one line. The actual "validated downstream by gsd-tools uat classify-coverage" descriptive mention now sits at line 416. Updated the allowlist entry's line number to match. - tests/task-command-router-resolve-content.test.cjs: the path-traversal test asserted the outside-project-scope diagnostic against the thrown ExitError's own .message. io.cts's error() (ADR-3889) writes its human-readable message to fd 2 via writeAllSync and then throws a bare `new ExitError(1)` with no message argument — by design, so the exception carries no duplicate text and the thrown ExitError's message defaults to "process exit 1" (cli-exit.cts's ExitError constructor). Root cause was the test, not the source: task-command-router.cjs's outside-project-scope rejection already calls error() correctly and the diagnostic text is genuinely emitted, just on fd 2, not on the exception. Fixed the test to capture fd-2 writes (mirroring tests/estimate-calibrate.test.cjs's runCalibrateExpectError and this same file's own captureStdout for fd 1) and assert against the captured stderr text instead of err.message. This was masked locally because a manual `node -e` sanity check that only inspects the caught exception's .message cannot see what the real node:test run actually failed on. Emitted-Drift-Ack-Growth: execute-plan.md — adds the ADR-3646 task-content-resolution bullet and checkpoint-exclusion caveat to the per-task execute loop; a real behavioral prose addition, not incidental bloat. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> * docs(#3970): backfill changeset PR number (pr:0 -> pr:4000) --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>	2026-08-28 13:17:04 -04:00
Tom Boucher	d24e22b156	enhance(#3912): gsd-tools declares outcomes, pinned at v1 (#3983) ... * enhance(#3912): gsd-tools declares outcomes, pinned at v1 ADR-3889 §4. Phase 6 already moved error()'s terminator onto the seam, so what remained was the declaration — and the pin that makes it invisible today. The census corrected two documented figures before any code changed. ERROR_REASON has exactly 25 members (the ADR and epic were right; an earlier note of mine claiming 23 was wrong and is corrected). And output({error}) is **64 sites across 9 files, not the 60 ADR-2980 ratified** — the module shape holds but the total drifted +4: frontmatter 7 not 6, phase 4 not 2, roadmap 3 not 2. That matters because this phase's criterion demands the pin be asserted over the enumerated population rather than sampled; asserting over a stale 60 would leave four sites unpinned while claiming full coverage, which is the shape of failure this epic exists to remove. The issue does not state the fact that shapes the design: output() never touches the exit code. Confirmed by reading it — it writes fd 1 and returns. So a declared outcome for those 64 sites had nowhere to be READ. The mapping was never the work; wiring somewhere for the declaration to land was. The seam already existed twice over. cli-exit.cts holds two globalThis-Symbol cells, each because the module is emitted to three locations and a module-level `let` would let instances disagree, and runMain already maps a code returned by main(). A third cell inherits that solution. output() records DEGRADED for any {error} payload — key-order agnostic, which is exactly why the "42 sites" figure undercounts — and runMain projects the cell only when main() returns nothing, so an explicit return still wins. error() maps its reason through a table over the closed 25-member enum, leaving all 278 call sites untouched; 226 of them pass no reason at all. The version gate lives in error(), NOT in projectOutcome: registered names are version-invariant there, so mapping a reason straight through would make USAGE project to 64 under v1 and break the pin on its first line. projectOutcome is left exactly as Phase 2 shipped it, DEGRADED's 0/80 asymmetry included. Proven rather than asserted. v1 is byte-identical across three real CLI paths — config-get plain, config-get --json-errors, and an output({error}) path — matching exit code and exact bytes against the pre-change build. Under GSD_EXIT_CONTRACT=v2 the same commands now exit 66 (CONFIG_KEY_NOT_FOUND -> NO_INPUT) and 80 (DEGRADED), both looked up through the registry. An anti-vacuity test pins that v1 and v2 genuinely differ for at least one reason, because without it a mapping where everything projects to 1 under both versions would satisfy every other assertion and the declaration would be theatre. A1 iterates all 25 enum members and A3 asserts over the measured 64-site population, so a 26th reason or a 65th site fails until it is given a mapping — the drift guard this phase needs, given ADR-2980's own count had drifted +4 unnoticed. Verification runs on the remote runner. Refs #3912 * fix(#3912): the outcome cell must never lower an exit code The remote run caught a fail-open that this phase introduced, in the phase whose entire purpose is removing fail-opens. `state validate --strict` on a missing STATE.md exited **0** where it must exit 1. Mechanism: `runMain` projected the pending outcome whenever `main()` returned void, and under v1 DEGRADED projects to 0 — so a `process.exitCode` already set non-zero by the command was clobbered down to success. Confirmed live against a fixture, before and after. This refutes a review conclusion recorded earlier in this phase, that the cell was "fail-closed and can never mask a failure as success". It could, and did. Recording that plainly so the assumption is not repeated: the cell's danger was never only that it might add a failure — it was that projecting it unconditionally overwrites whatever decision came before. Projection is now guarded: it may set a code only when none is set, and an already-non-zero exit code always wins. The full precedence — explicit `main()` return, then an existing non-zero exitCode, then the declared outcome — is written at the projection site. A regression test drives a void return with a pre-set non-zero code and a pending DEGRADED, and fails against the pre-fix build. The second failure was my test encoding the wrong contract, not a code defect. It asserted `output({found:false, error: undefined})` records DEGRADED because the KEY is present. `JSON.stringify` drops undefined, so the payload the user receives is `{"found":false}` — carrying no error at all, and calling that degraded would hand back exit 80 under v2 for output that reads as clean. The discriminator is a serializable error VALUE, not key presence. The test now pins `{error: undefined}` as explicitly NOT degraded, and the design doc's wording is tightened to match. Verification runs on the remote runner. Refs #3912 * docs(#3912): the versioned exit contract, and a flag defect the docs found Diataxis pass for Phase 8, plus a real fix that only surfaced because writing the how-to meant running its own examples. The docs. ADR-2980's "Revisit if" clause asked for exactly the versioned projection this phase provides, so it gets an amendment naming #3912 / ADR-3889 section 4 as that boundary: v1 stays 0 byte-for-byte, v2 projects DEGRADED to 80. The amendment also records the count drift rather than restating a stale figure — the ADR ratified 60 output({error}) sites in 9 modules; the AST-measured population is 64 across the same 9 (frontmatter 7 not 6, phase 4 not 2, roadmap 3 not 2). The pin is asserted over the enumerated 64. json-errors.md gains the outcome-declaration reference, including the precedence order a review pass got wrong and the suite refuted: an explicit main() return, then an already-set non-zero process.exitCode, then the declared outcome. Projection may only ever set a code, never lower one. A how-to is owed here and is written, not skipped. Under v1 nothing changes, so the audience is an operator opting into v2 and needing to know what the codes mean for a CI gate — a migration, which is how-to shaped. It covers turning v2 on, the code table, why 80 is "ran and reported a condition" rather than a crash, and how to split a gate that treats any non-zero as fatal. No tutorial: there is no new entry point to learn, and under the default contract a reader would be walked through observing nothing. The defect. Running the how-to's own Step 1 example returned $ gsd-tools --exit-contract=v2 state validate --strict Error: Unknown command: --exit-contract=v2 (exit 64) while the same flag trailing the subcommand worked and exited 80. The flag half-worked, by argv position. resolveContractVersion scans argv non-destructively, so the token survived into the dispatcher, which treats argv[2] as the command name. --json-errors had already solved precisely this at gsd-tools.cjs:4455, under a comment naming the hazard verbatim: "The argv splice must happen here too, otherwise the dispatcher below sees --json-errors as an unknown command." The later flag never got the same treatment. Fixed rather than documented around: the version is resolved first — which memoizes the cell and makes an invalid value throw early — and then every --exit-contract= token is spliced out of the dispatcher's argv copy. --exit-contract is now listed in TOP_LEVEL_USAGE, where it never was. The regression test pins leading position, trailing position, agreement between the two, and a loud failure on v3 rather than a silent fall back to v1. Neither review engine would have caught this: the defect is invisible in the diff, because the diff does not touch argv handling. It surfaced only from running the documentation's own example. Writing a how-to is an execution pass. Verification runs on the remote runner. Refs #3912 * fix(#3912): the flag splice has to run before the run-with-timeout return An isolated review of the previous commit found that the fix did not deliver what it claimed, and that two of its own tests were weak. All three findings reproduced by execution before any change was made. The fix was placed below a return. main() intercepts `run-with-timeout` at gsd-tools.cjs:4436 and returns from there — above both the --json-errors block and the --exit-contract splice added in the previous commit. So the flag still died in leading position for that one command: $ gsd-tools --exit-contract=v2 run-with-timeout 5 -- node -e "..." Error: Unknown command: run-with-timeout (exit 64, child never ran) The previous commit message and the test's describe-block both claimed position-independence unconditionally. That was an overclaim, not a gap left open, and it is the part worth naming: the fix was verified by hand on the commands I happened to think of, and `run-with-timeout` returns before the code I was verifying. Both global-flag blocks now run above the interception, with a comment naming it so a later edit cannot slide them back down. Moving --json-errors up fixes the identical pre-existing bug for that flag, verified failing beforehand (exit 1, sdk_unknown_command). Fixing the sibling is deliberate: same defect, same block, and a known-broken twin next to a fixed one is not a resting state. Two tests were not pulling their weight. The invalid-value test was vacuous — it passed against the pre-fix build, because `--exit-contract=v3` already exited 1 there and already printed the resolve error lazily through error() -> getContractVersion. Both its assertions held before the fix, so it pinned nothing. The real discriminator is that the pre-fix build emits BOTH "Unknown command: --exit-contract=v3" and the resolve error, while the fixed build emits only the latter; the test now asserts that absence. The leading-position and leading==trailing tests asserted proxies — "not 64", "no Unknown command", "the two agree" — none of which pin a value, and all of which would survive both positions being identically broken. With a .planning directory and no STATE.md, state-snapshot exits exactly 80 under v2 and 0 under v1 in both positions. Those numbers are pinned now. The multi-token case the descending splice loop exists for is covered too, and run-with-timeout has regression tests for both flags. The lesson is narrower than "test more". Hand-verifying the production behavior does not verify that the test would have caught its absence. The pre-fix binary has to be run against the test's own assertions. Investigated and deliberately not changed: splicing before --cwd parsing degrades one diagnostic from "Missing value for --cwd" to "Invalid --cwd: <path>", but that is pre-existing — verified on the pre-fix build via --json-errors, which already did it. This change joins the pattern rather than creating it, and both forms exit 64 on malformed input either way. Verification runs on the remote runner. Refs #3912 * chore(#3912): backfill changeset pr numbers to 3983 * test(#3912): pin the reason-table invariant as set equality, not a count A graph-backed review flagged the unchecked lookup in expectedErrorCode3912. Investigated by execution: the drift guard DOES hold — for an unmapped reason under v2 the production error() yields 1 while the table yields undefined, so the assertion fails. Not a correctness defect, and deliberately NOT made tolerant, since a tolerant lookup would destroy the guard. Two real problems remained. The guard asserted the wrong invariant: it counted the TABLE's keys at 25 rather than checking they match the ENUM's values, so a renamed member keeps the count at 25 and slips past, and a 26th member leaves the table at 25 and slips past too. Both were then caught only indirectly, by an undefined mismatch producing 'must exit undefined'. It is now a sorted set equality, so the failure names the specific missing or extra reason. And the comment above it described a '?? FAIL' fallback that does not exist anywhere in the function. It now states what the code actually does, verified by running it rather than by reading it. Refs #3912 --------- Co-authored-by: sim <sim@local>	2026-08-28 08:09:05 -04:00
Tom Boucher	d98b55562c	enhance(#3910): the raw terminator is banned by construction (#3980) ... * enhance(#3910): move the last src/ terminators onto the seam Phase 6 bans the raw terminator by construction, which it cannot do while violations stand. A census found 12 sites the rule would flag; nine of the ten unsanctioned ones were owned by no phase of the epic at all — a coverage hole in the decomposition, since P0-P2 are infra, P3 the gate modules, P4 the scanners, P5 the fragments, P7 the hooks, P8 io.cts, and P6 itself only adds the rule. `src/**/*.cts` now holds exactly 2 raw exits, both inside `terminateNow`, the single sanctioned site. `io.cts`'s `error()` is the interesting one. It was first called substantive on "dozens of callers, contract risk" — asserted, not measured, and the measurement refuted it: 289 call sites, zero inside a try whose catch would swallow a throw. The real obstacle was structural instead: `terminateNow` cannot emit exit 1, because ADR-3889 §1 makes 0 and 1 unallocatable and `nameForExitCode(1)` throws. So the only route is `ExitError` under `runMain`, which sets exitCode and writes stderr only when the error carries a user message — keeping the existing stderr write and throwing a message-less ExitError is observably identical. That census was still too narrow, and running the CLI proved it. It asked whether the CALL sits in a try/catch; the two regressions that surfaced were interceptors elsewhere on the stack: - `command-routing-hub.cts`'s `dispatch()` swallowed the ExitError into a HandlerFailure, so the caller emitted a duplicated, wrong stderr line on every Hub-routed path. It now rethrows ExitError explicitly — the same shape `gsd-tools.cjs` already used at two dispatch sites, so this follows an established idiom rather than inventing one. - the profile-pipeline router's deliberately un-awaited `.catch(e => error(...))` turned an ExitError rejection into an uncaught exception; it now mirrors runMain's handling. `edge-probe` and `ui-consideration-probe` gained `runMain` wrappers because probe-core's new throwing default would otherwise have escaped them. A follow-up sweep of every dispatcher — 19 command routers, the Hub, the gsd-tools dispatch seams — found no further swallowing catch. The admitted bound: ~1260 non-rethrowing catches repo-wide were scanned structurally but not individually classified. Both real regressions were found by execution, not by reading, so the suite is the detector that matters here. `gsd-tools.cjs:253` stays a raw exit deliberately: it is the ensureRuntimeBuild bootstrap, which runs before cli-exit is required, so the seam does not yet exist. It needs a second allowlist entry, which means #3910's "single allowlist entry" criterion is unachievable as written. Verification runs on the remote runner. Refs #3910 * enhance(#3910): ban the raw terminator by construction Adds local/require-registered-exit and registers it on all four globs: src/**/*.cts, scripts/**/*.cjs, hooks/**/*.js, gsd-core/bin/**/*.cjs. Registering on the .cts glob is load-bearing, not redundant — the emitted .cjs mirrors are globally eslint-ignored, so a rule registered only on the emitted globs is blind to the sources. That is the #3496 lesson, and it is how the previous guard became invisible: n/no-process-exit was 'error' in one block yet fired zero times on all three surfaces that mattered. The dead n/no-process-exit: 'off' block for hooks is deleted in the same PR. Phase 7 migrated every hook, so the exemption now protects nothing. Two allowlist entries, not the one #3910 anticipated. terminateNow's body is detected STRUCTURALLY — a process.exit lexically inside a function of that name — rather than by a path and line number that rots. The second is gsd-tools.cjs's ensureRuntimeBuild bootstrap, an inline disable with its reason at the call site: it runs before ./lib/cli-exit.cjs is required, so the seam does not exist yet and no migration is possible. #3910's 'single allowlist entry' criterion is therefore unachievable as written, and is amended with the measurement rather than quietly missed. The rule is proven able to FAIL, per glob: four positive controls, one for each registered glob. A guard that cannot be shown to fire is not a guard. Four matching negative controls pin process.exitCode as never-flagged — conflating it with process.exit is what inflated this epic's original census 2x. An allowlist case and a near-miss (same shape, different function name) fix the structural detection in place. Verification runs on the remote runner. Refs #3910 * fix(#3910): stop the detached catch from throwing, and scope the allowlist Review findings, one of them a regression the previous fix introduced. _handlePipelineRejection called error() from inside a DETACHED .catch(). error() now throws, so that throw became an unhandled promise rejection — and on Node >=15 with --unhandled-rejections=throw, Node dumps a raw stack trace with absolute paths on top of the clean Error: line. That was impossible before this branch, because process.exit(1) terminated synchronously before any rejection machinery could observe it. The handler now writes byte-identical stderr itself, in both plain and --json-errors form, and sets exitCode in place. This was the THIRD interceptor found, and like the first two it surfaced by running the CLI rather than by reading code. The rule's terminateNow allowlist had no path constraint, so any function anywhere named terminateNow across all four globs inherited it. It now requires the structural nesting check AND a cli-exit.cts basename — still no line numbers to rot. The four per-glob positive controls only varied a filename inside RuleTester, which never resolves eslint.config.mjs. Since the rule is filename-agnostic, all four exercised identical logic and none proved the rule was WIRED — this epic's own failure mode. A registration test now asserts the rule resolves for a real path in each glob, and it is proven able to fail: removing one glob's registration flips the resolved value from [2] to undefined. Three evasions the rule cannot catch (computed member, aliasing, .call/.apply) are documented in its header and pinned by tests, labelled as known limits rather than endorsed, so a future change that starts catching them is a deliberate diff. Refs #3910 * docs(#3910): document the raw-terminator ban Reference and Explanation via a new docs/features fragment (FEATURES.md is generated from it, not hand-edited). How-To: docs/how-to/resolve-a-raw-terminator-finding.md, indexed from docs/README.md — a contributor whose code trips the rule picks among three replacements by surface (runMain/ExitError for a CLI path, terminateNow for a hook, process.exitCode where the process should drain), and needs to know why process.exitCode is correct and never flagged, since conflating the two is what inflated this epic's original census 2x. The page also names the three patterns the rule cannot catch and says plainly that using one to dodge it is a review finding, not a fix — documenting them without that sentence would read as a sanctioned workaround. docs/INVENTORY.md deliberately untouched: eslint-rules/ is not a tracked family in the manifest (verified — a regen produced a zero diff), so a hand-written row would desync the table from the family it claims to belong to. Refs #3910 * fix(#3910): a catch that sniffs the message swallows an ExitError The remote run returned 41 failures, and one of them was a live production regression rather than a test artifact. `cmdMilestoneComplete`'s unstarted-phase guard re-threw only when `e.message.startsWith('Cannot mark milestone complete:')`. `error()` used to `process.exit(1)`, uncatchable, so the guard always fired. It now throws an ExitError carrying no message, the string test fails, and the ExitError was silently swallowed — the guard stopped blocking milestone completion entirely. Proven against the real CLI: pre-fix, a milestone with an unstarted phase archived at exit 0; post-fix it is blocked at exit 1 with the intended message. That is a guard that silently stopped guarding, which is this epic's thesis appearing inside the phase meant to enforce it. Worth stating plainly: an earlier census DID examine this site, saw a `throw e`, and classified it as rethrowing. It was wrong — the rethrow is conditional, and a conditional rethrow on an inspected message is indistinguishable from an unconditional one unless you read the predicate. So the class was swept rather than patched where it was tripped over. An AST census of every CatchClause across src/, gsd-core/bin/ and scripts/ found 38 conditional rethrows. Two more had the same defect and are fixed the same way: `config.cts`'s `'No config.json'` sniff and `gsd-tools.cjs`'s `e.name === 'WindowsError'`. The remaining 25 are provably unreachable — every one wraps a bare fs, YAML, manifest-require or git-exec primitive that cannot throw ExitError — and two were scanner false positives, both explained. Each fix is an unconditional `instanceof ExitError` rethrow placed BEFORE any inspection, matching the idiom command-routing-hub and gsd-tools already used. Residual bound, stated rather than implied: zero known-reachable unfixed sites, contingent only on error() never later being called inside one of those 25 primitive try blocks. The remaining failures were harness artifacts, and the harnesses were corrected to the new contract rather than the assertions weakened. Tests that mocked `process.exit` to observe termination now catch ExitError and assert its code; tests parsing stderr as a single JSON object still assert exactly that, with their ad-hoc `node -e` scripts wrapped in runMain so it is true. milestone and phase-resolution-parity needed no test change — they were correctly written against the real bug and are what caught it. Verification runs on the remote runner. Refs #3910 * chore(#3910): backfill the changeset PR number Also reframes the fragment to lead with the user-visible change — the milestone guard blocking again — rather than the narrowest of the three fixes. Refs #3910 --------- Co-authored-by: sim <sim@local>	2026-08-28 03:15:39 -04:00
Tom Boucher	15af0f5536	enhance(#3951): B6+B7 — widen two unreachable lint rules and make the guard ledger true (#3965) ... * fix(#3951): two lint rules that could not reach the code they govern B6 names two widenings. Measuring them first turned up a defect the criterion did not know about, and refuted the reason it gave for one of them. 1. no-adhoc-markdown-parsing self-gates on its own filename. Lines 107-110 short-circuit create() to {} unless the path matches /(?:^|\/)src\/[^/]+\.cts$/. B6 says to widen the files: glob in eslint.config.mjs - but doing only that ships an INERT rule, because the gate still returns {} for every new path. Both halves have to change, and the gate is the load-bearing one. That same regex hides a live hole: [^/]+ is FLAT-ONLY, so it requires the file to sit directly in src/. The registered glob is src/**/*.cts, which includes subdirectories. 28 .cts files - health-diagnostic-rules/ (10), installer-migrations/ (11), observability/ (3), host-integration-adapters/ (2), vendor/ (2) - are inside the registered glob and silently skipped. Measured with the gate neutralized: 0 violations there today. The hole is hiding nothing right now, and is fixed anyway, because "no violations today" is not a property that keeps holding. The fix is not invented: require-subprocess-timeout.cjs:196 already carries the correct form of this guard, /(?:^|\/)src\/.*\.cts$/ with .*, one directory over. Checked the other 21 rules for the same bug - no-adhoc-regex-escape and no-private-binary-resolution short-circuit only to exempt their own seam file, which is the right shape, and no-crlf-fragile-split has no filename gate at all. This bug is unique to the one rule. 2. no-adhoc-regex-escape could not see the shape that actually occurs. Line 396 gated the whole UNSAFE-NEW-REGEXP arm on arg.type === 'Identifier'. Every check below it - the _SOURCE provenance check, the isSoleReturnOfOwnParameter shape - lives inside that branch, so new RegExp(obj['key']) and new RegExp(cfg.pattern) were never examined at all. Runtime data arrives as a property access far more often than as a bare identifier, which is exactly why this rule never fired on the #3477 ReDoS. Widened to MemberExpression, measured by AST walk across all five registered blocks rather than by grep. 27 sites, zero TSAsExpression: 18 safe new RegExp(X.source, flags) -> exempted, keyed strictly on the PROPERTY being `source`, never on the object. Keying on the object would wave through X.anything and buy nothing. B6 estimated ~10; that was an undercount. 3 _SOURCE-suffixed constants reached through a required module namespace (phaseId.BRACKET_PHASE_TOKEN_SOURCE) -> the same provenance-exempt class the rule already recognizes for bare identifiers, extended to reach them. Without this the widening produces 3 false flags. 6 real findings -> marked, each a test extracting a pattern from a shipped file at test time, where the runtime contract IS the product. Deliberately the NARROW MemberExpression form. The rule's own isSoleReturnOfOwnParameter doc comment records that an earlier broad "any non-literal identifier" heuristic produced ~25 false positives and was rejected; a re-run of the census after this change flags exactly the 6 above and nothing else. Verified by execution, not by reading: the gate now accepts src/<subdir>/x.cts, still accepts flat src/x.cts, and still exempts paths outside src/ - each pinned by a test proven to fail against the old regex. build:lib, lint and lint:ci all exit 0. Refs #3951 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3951): give no-adhoc-markdown-parsing its reach, and fix the 80 parses it finds The rule self-gates on filename AND is registered on one glob, so widening either half alone is inert. Both move here: the gate now accepts tests/**/*.cjs and scripts/**/*.cjs alongside src/**/*.cts, and eslint.config.mjs registers it on the same two. A test pins that the gate and the registration AGREE, in both directions. The original defect was a gate narrower than its registration; the failure mode of this fix is a gate wider than its registration. Both are silent, so the test asserts the pair rather than either half. 80 violations across 43 files, all in tests/, zero in scripts/. 70 are routed through the existing seams - scanFencedBlocks, collectSection, stripFencedCode, tokenizeHeadings from markdown-sectionizer; splitTableRow, parseMarkdownTable, findTableWithColumns from markdown-table. Headerless STATE.md tables use splitTableRow per line, because parseMarkdownTable needs a real delimiter row. 10 are suppressed, 12.5%, well under the third that would have meant the rule is mis-scoped for tests/ rather than the tests carrying debt. Each names its reason: three regression guards (#3873 / bug-#21) are deliberately independent of the generator's own fence handling, and routing them through the seam would have them test the generator against itself; one is a negative-text probe that extracts nothing; six are a shell-pipe-to-jq detector whose regex coincidentally matches the table fingerprint and is not markdown parsing at all. All ten sit in tests whose subject is .md content, which is normally a reason to prefer the seam. The marker used is allow-adhoc-markdown, distinct from no-source-grep's allow-test-rule, and lint:ci's lint-allow-test-rule-refs reports the same 280/280 unverified count as before - checked rather than assumed, because those two markers are easy to conflate. The widening earned its keep immediately: it found a test that passed for the wrong reason. tests/config-field-docs.test.cjs asserted notEqual(<cell>, '600') against the TYPE column instead of the DEFAULT column. notEqual('number', '600') is true forever, so the guard against workflow.subagent_timeout regressing to the old seconds default could never fire. docs/CONFIGURATION.md:434 is `| workflow.subagent_timeout | number | 300000 | ... |`, so the default is cell index 2; the assertion is now row-scoped through splitTableRow and reads 300000. That is the argument for the widening in one case: the violation was invisible to lint, the suite was green, and the assertion was vacuous. A rule that cannot reach a file cannot tell you the file is lying. Not fixed here, and recorded rather than assumed: #3426/#3239 are NOT reachable by this widening. tests/package-legitimacy-gate.test.cjs yields zero violations even with the gate bypassed - its hand-rolled scans are real, but built from line filters and split('|') rather than the regex-literal fingerprints this rule detects. They need new detectors. The epic assumed a wider glob would catch them. build:lib, lint and lint:ci all exit 0; the post-fix census across tests/** and scripts/** is 0 violations. Refs #3951 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3951): B7 — and #3356's defects were still live in the code B7 asks that each closed child be driven fail-first with a behavioral identity test at the CONSUMER's output. Four of eleven children had no test citing their issue number. Auditing them by BEHAVIOR rather than by number-grep changed the answer for three of the four. #3364 and #2540 — traceability only. Both were implemented by #3941 and their consumer-output tests exist and were shown failing-first; neither cited its originating issue, so an audit that greps for the number reports them uncovered. Tagged the specific asserting test in each file, following the citation form those files already use. #3372 — covered, but only at helper level, and the triage narrowed it. Of the four commands the issue names, only estimate-cli's collectCalibrationSamples actually enumerates phase dirs from disk; smart-entry, audit and roadmap-upgrade derive from ROADMAP/body text and never reach the sentinel path, so they are benign by construction and were left alone rather than "fixed" into churn. The existing #3882 rows asserted the helper's return value. Added a consumer-output test driving `query estimate-calibrate` and asserting sample_count and the persisted document. RED proof: reverted collectCalibrationSamples to a raw readdirSync and ran the real CLI - sample_count 3, sentinel leaked; restored - sample_count 2. #3356 — NOT covered, and BOTH halves of the defect were still live in source. The issue is closed; the bug was not fixed. Fixed here rather than writing tests that document a bug as correct. Defect 1, the contradicted row. quick.md:627 claimed `quick-tasks-append` performs "the equivalent write" to the Step 7c row. It did not: the `#` cell was a positional ordinal and `Directory` read `—`, because the route had no way to receive a quick id or task directory. Added OPTIONAL `--quick-id` / `--slug` / `--directory`. A caller with neither - fast.md, the original #2133 caller - omits them and gets the byte-identical prior row, so nothing existing changes. A caller that HAS a real id and directory now gets the canonical row quick.md:632 renders. The false-equivalence sentence itself is corrected rather than left to mislead the next reader. Defect 2, the forced re-derive. The route called readModifyWriteStateMd with no options, so a body-only append to the Quick Tasks table triggered a full re-derive of the disk-derived progress.* frontmatter. Every other body-only writer passes { resync: false } - src/state.cts's own docstring prescribes it - and this route was the lone outlier. RED proof: reverted the option, seeded a project with 2 real phase dirs and a curated total_phases of 25, ran quick-tasks-append; total_phases collapsed to 2. Restored; it stayed 25. That second one is the shape this epic exists to close: a silent write that replaces curated state with a re-derivation nobody asked for, exit 0 throughout. build:lib, lint and lint:ci all exit 0. Refs #3951 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(#3951): amend B6's ledger to what was measured, and document the new flags The ADR gains a ledger amendment in its own correction style - the sixth wrong premise it records, found the same way as the other five, by measuring before building. B6 says the net guard count must fall. It rose: 62 -> 69, +7, measured from the epic's filing commit to origin/next. The attribution is the point, though. Five of the seven came from PRs unrelated to this epic, one was added by a phase of it, and the epic did retire something sub-file - #3884 removed a detector with an explicit "net: -1 detector, 0 added" ledger. Every named casualty is load-bearing, two already carry retractions in this same document, and a sweep of all 22 rules plus every scripts/lint-* found no provably dead guard. There is no honest way to make the count fall; forcing it would trade coverage for a number, which is the Goodhart outcome Decision 6 exists to prevent. The amendment also records that B6's own prescribed fix for one widening was inert. no-adhoc-markdown-parsing self-gates on its filename, so widening only the files: glob - which is what the criterion says to do - ships a rule that still returns {} for every new path. And #3426/#3239 are not reachable by that widening at all; their scans use line filters and split('|'), not the regex fingerprints the rule detects. The roster row tracked them against the wrong mechanism. Three roster rows updated from aspiration to fact: the two widenings are DONE with their measured counts, and lint-phase-enumeration-drift is marked RETAINED rather than "expected casualty - verify before retiring", because Phase 5 verified it and kept it. The rule Decision 6 should carry forward is stated plainly: a guard ledger is a claim about COVERAGE, not about COUNT. "Net count must fall" is measurable and wrong. "Every guard is reachable, and each retirement names what makes its defect unrepresentable" is the property that was actually wanted. CLI-TOOLS.md documents the optional --quick-id/--slug/--directory flags and says plainly that omitting them keeps the pre-#3356 row byte-identical, plus that the append no longer re-derives progress frontmatter. New features fragment (id 3951); FEATURES.md regenerated rather than hand-edited. Changeset is Changed, pr:0 pending backfill. Refs #3951 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test(#3951): correct four rows that pinned the lint rule's old narrow reach The remote suite came back RED with 5 failures, all in tests/eslint-rules.test.cjs. They are stale tests, not a regression: four rows assert that no-adhoc-markdown-parsing is inert outside src/*.cts, which is exactly the contract this deliverable changes. Confirmed by reading rather than inferred from the names - the row at :1981 used filename: 'tests/some.test.cjs' and filename: 'scripts/helper.cjs', the two roots the rule now covers on purpose. Worth recording WHY local gates missed this. npm run lint and lint:ci were green, and the touched test files passed standalone. Lint only reports violations in real files; these rows assert the rule's REACH using synthetic RuleTester filenames, so nothing but the full suite could see them. Local green on a rule change says nothing about the rule's own tests. Each row is rewritten with BOTH halves rather than flipped from valid to invalid: - the same fingerprint under tests/ or scripts/ is now flagged, with the right messageId - the negative space is preserved - the same fingerprint under a path outside all three roots (gsd-core/bin/lib/foo.cjs) is still NOT flagged The second half is the one that matters. Without it the rule has no boundary and nothing would catch an over-wide gate later, which is the mirror image of the bug this deliverable just fixed. Each row is renamed to state the current contract; the old names said "non-src/*.cts ... is not flagged" and would have been actively misleading once the bodies changed. Proven to test the widening rather than restate it: every flagged half was run against HEAD~2's pre-widening rule and does NOT fire there, then against the current rule and does. 12/12 on that probe; the full file is 178/178. Swept for the same staleness elsewhere and found none. require-subprocess-timeout's own "inert outside src/*.cts" row is untouched - that rule's gate was not widened here - and no-adhoc-regex-escape's test file already carries correctly-targeted rows. Refs #3951 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test(#3951): acknowledge the quick.md growth the attribution guard reported The full suite came back RED with one failure, and it is mine: 1 file(s) grew without an acknowledgment: quick.md grew 364 bytes gsd-core/workflows/quick.md is runtime-loaded emitted content, so correcting its false 'performs the equivalent write' claim trips emitted-attribution by construction. This is the acknowledgment, not a workaround - there is nothing to regenerate. The fragment names ONE path, which is the only one the guard reported. The four spent acknowledgments it also listed (audit-uat, plan-phase, progress, review) belong to other fragments whose ripple the base already absorbs; they are inert, not failures, and are deliberately NOT copied here - naming paths I did not change would make this record false in the other direction. Byte figure corrected before committing: the guard reported 37220 -> 37584 (+364), but origin/next has since moved and quick.md is 37232 there now, so the measured delta is +352. The reason text says so and names the base as a moving figure rather than pinning a number that is already stale. Refs #3951 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test(#3951): move the quick.md growth ack to a trailer, delete the obsolete fragment The acknowledgment mechanism changed under this branch. Merging next brought in the redesign - it also deleted .github/workflows/ack-fragment-sweep.yml, which was in the merge status and which I did not register at the time - and the guard now says so directly: Add a trailer to a commit in this PR (never a new file). Emitted-Drift-Ack-Growth: quick.md - <why this growth is deliberate> So tests/emitted-drift-acks/3951-quick-append-equivalence.json is obsolete on arrival. A fragment file is no longer read by anything, and leaving it would be a dead record that looks like an active one. It is deleted here rather than kept "just in case". The byte figure moved again with the merge: 37232 -> 37596, +364. The earlier fragment said +352, measured before the merge auto-merged quick.md itself. The trailer carries no number, which is the better design - the figure was stale twice in two attempts. Refs #3951 Emitted-Drift-Ack-Growth: quick.md — #3356/#3951 replaces a false claim with an accurate one. Line 627 said the `quick-tasks-append` shortcut "performs the equivalent write" to the Step 7c row rendered above it; it did not, and that was the documented half of #3356 — with no quick id or task directory the route emitted a positional ordinal in `#` and an em-dash in `Directory`, a visibly different row. The corrected sentence has to carry three facts the original elided: what the shortcut actually writes when it has neither input, that this is honest behavior for its real caller (`fast.md`, which has neither), and how a caller with both now gets the byte-identical canonical row via the new optional `--quick-id`/`--slug`/`--directory` flags. Prose is the product here — an executing agent reads this line to decide whether the shortcut is safe for its case, and a shorter correction would either drop the flags (leaving the reader unable to act on the fix) or drop the limitation (recreating the false claim in gentler words). Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#3951): backfill changeset pr number Refs #3951 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-27 23:10:49 -04:00
Tom Boucher	2ea5efc151	enhance(#3911): hooks declare their crash policy (#3960) ... * enhance(#3911): give hooks an exit seam that needs no build ADR-3889 Phase 7 foundation. The 19 shipped enforcement hooks hold 91 of the epic's 128 terminators and cannot reach `terminateNow` today. The obvious route — requiring `gsd-core/bin/lib/cli-exit.cjs`, as gsd-agent-isolation-guard.js already does for two other modules — is rejected. That precedent carries its own warning (#3582): those files are tsc output, gitignored and absent on a raw plugin-marketplace or git-clone install, so the hook must first call ensureRuntimeBuild() to self-heal. Making the module a hook needs IN ORDER TO TERMINATE depend on a build inverts the dependency, and its failure mode is precisely the fail-open this phase exists to remove: a guard that cannot terminate cannot deny. `lint-hooks-runtime-build-seam` already encodes that concern, and Design B would have had to add an ensureRuntimeBuild() call to all 19 hooks to satisfy it. So `hooks/lib/` becomes a third emit location for cli-exit and a fifth for the registry, preserving the invariant `src/cli-exit.cts`'s own header states: it imports nothing but node:fs and its sibling registry, and the generator dual-emits that sibling alongside each copy so a relative require resolves next to whichever copy loaded it. Shipping needed no change — build-hooks.js already declares HOOKS_SUBDIRS_TO_COPY = ['lib']. Proven, not asserted: the two files are copied into an otherwise-empty tmpdir and a child process requires them and terminates — PASS exits 0, HOOK_DENY exits 2 with the payload on both stdout and stderr. That test fails the moment the hooks copy gains a require reaching outside hooks/lib/. Also fixed inline: the registry's fifth target let any `--write` test overwrite the real committed hooks/lib/exit-code-registry.js, because the test helper derived only three of the other output paths. It now redirects all five, and a regression test asserts every committed artifact is byte-identical after a redirected write. Install-tree goldens pick up the two new shipped paths across 11 runtimes — insertions only, no removals. lint:ci was green while they were stale, so this was found by regenerating rather than by a gate. Verification runs on the remote runner. Refs #3911 * enhance(#3911): declare a crash policy, and migrate the write guard Adds `hooks/lib/hook-exit.js` — the hook-facing vocabulary over `terminateNow`, hand-written because the cli-exit copy beside it is generated: allow(payload) exit 0 deny(payload, stderr?) exit 2 crash(onCrash, payload) whichever the hook DECLARED `crash()` takes the policy as a required argument with no default, which is the whole mechanism: fail-open by accident stops being expressible. A hook must name ALLOW or DENY at the call site, and an unrecognized value terminates INTERNAL rather than guessing. Fail-open stays legal; fail-open by omission does not. `gsd-write-guard.js` is the first hook migrated, all 12 sites, and it exposed a gap in the seam. `terminateNow`'s doc comment justified its fd-2 write by citing this hook's `emitBlock` — but modeled it as sending the same bytes to both streams, when `emitBlock` actually sends full JSON to stdout and only the bare `reason` string to stderr, because Kimi's hook bus feeds stderr verbatim back to the model. Migrating as written would have turned a readable sentence into a JSON blob for Kimi-backed agents. #3911 requires both "all 19 hooks terminate through terminateNow" and "no hook's effective default changes". Those are jointly satisfiable only by teaching the seam to carry a distinct stderr payload, so `terminateNow` gains an optional third argument: omitted, behavior is byte-for-byte what it was; a string is written raw, which is exactly the Kimi case. The doc comment's inaccurate claim about emitBlock is corrected in place. Proven rather than asserted: the pre-migration file is reconstructed from HEAD and driven with the same catastrophic-shrink payload as the migrated one — exit code, stdout and stderr all byte-identical. Verification runs on the remote runner. Refs #3911 * enhance(#3911): all 19 hooks terminate through the seam Migrates the remaining 18 enforcement hooks onto allow/deny/crash. An AST walk now reports zero `process.exit(` call sites across every `hooks/*.js` — down from the 91 the census measured. Each hook with an outer catch declares its policy once, at module top, with the reason that policy is right for that specific guard: a read guard that cannot scan must not block the read; a statusline that renders every prompt must degrade rather than crash; an injection scanner must not retroactively block a result already returned. Those sentences are the deliverable — they are what turns fail-open-by-accident into fail-open-on-purpose. No hook's effective default changed. Wiring exposed two defects, both fixed here rather than noted. A SECOND stdout/stderr-splitting site turned up in `gsd-workflow-guard.js`'s `emitForceAddBlock`, matching the pattern already known from the write guard — full JSON to stdout, bare reason to stderr for the Kimi bus. It uses the `stderrPayload` argument added in the previous commit, which is now carrying its second real caller rather than one special case. More seriously, `terminateNow` emitted both streams inside ONE try, so a payload that failed to serialize aborted before the stderr write ever ran. The two windsurf guards write nothing to stdout on a block and only a reason string to stderr, so `deny(undefined, reason)` exited 2 with EMPTY stderr — a deny that silently loses its reason, which is the exact "fails with success" class this epic exists to close. The streams are now emitted independently, each with its own guard, and `undefined` means "nothing to write for this stream" rather than an error. Regression tests inject a throwing write on one fd and assert the other still receives its payload; they fail against the single-try version. Byte-identity was proven per hook, not assumed: each pre-change file is reconstructed from HEAD and driven side by side with the migrated one across its normal path, its deny path, malformed stdin and empty stdin — exit code, stdout and stderr compared. Verification runs on the remote runner. Refs #3911 * enhance(#3911): harden the three shell hooks, and pin every hook's policy `gsd-phase-boundary.sh`, `gsd-session-state.sh` and `gsd-validate-commit.sh` gain `set -euo pipefail`. The expected hazard did not materialize, and that is worth recording: every intentionally-non-zero command in all three is already the condition of an `if`/`elif`, which `set -e` never fires on, and none of them reads a possibly-unset variable or pipes through a grep that may legitimately match nothing. No `|| true` guards were needed. Each hook was still checked command-by-command before the flags went in rather than after. Twenty-one before/after cases across the three hooks — disabled and enabled, planning and non-planning, missing STATE.md, malformed JSON, the Kimi payload shape, quoted and unquoted `-m`, valid and over-long Conventional Commits — all match on exit code, stdout and stderr. The hardening is shown to actually fire, not merely added: with a stubbed `node` that fails at the JSON-emit step, phase-boundary and session-state go from silently exiting 0 with empty stdout to failing visibly with the error surfaced. No such case could be constructed for `gsd-validate-commit.sh`, whose every statement already sits inside an if-condition — recorded as unproven rather than claimed. `tests/hooks-crash-policy.test.cjs` adds the per-hook coverage the issue asks for, table-driven over all 19 hooks rather than 76 hand-written cases: normal allow, deny where a deny path exists, crash-honors-the-declared-policy, and an unclosed-stdin case — the one `process.exitCode` structurally cannot serve. The deny assertions encode each hook's ACTUAL stream split rather than a uniform shape, since four of the six deliberately differ. A drift guard enumerates `hooks/*.js` and fails if a terminating hook is ever added without a row. Writing those tests surfaced two hooks that emit a block decision in their JSON body and exit 0. Both were checked rather than assumed, and neither is a fails-with-success: `gsd-read-injection-scanner.js` is PostToolUse, where the tool has already run and exit 2 has no meaning, and `gsd-cursor-subagent-start.js` follows Cursor's JSON-body protocol. They are deliberately left alone — a mechanical sweep to `deny()` would have broken exactly these two. Verification runs on the remote runner. Refs #3911 * fix(#3838): the commit validator says when it could not validate #3911 claims to subsume #3838. Measurement said otherwise, so this closes it for real rather than by assertion. `set -euo pipefail`, added earlier on this branch, does NOT fix #3838: bash exempts a command used as an `if` condition from `set -e`, and all three of the hook's swallow-and-pass sites are exactly that shape. Verified against the hardened hook with a node shim that fails only the classifier call — a non-conforming commit still exited 0 with empty stdout AND empty stderr, indistinguishable from "your commit conforms". That is the defect verbatim. All three sites named in #3838 now capture the real exit status instead of consuming it as a condition, and each distinguishes its genuine negative from "could not run": - the classifier: 0 = is a git commit, 1 = genuinely not one, anything else = could not classify. Its `node -e` now wraps the require and the call in try/catch and exits 3 on a throw, so a broken require chain can never be mistaken for `isGitSubcommand` legitimately returning false — which is the arm that matters, since `token-scanner.cjs` is a gitignored build artifact and a fresh checkout lands there. - the opt-in config read and the JSON command extraction get the same treatment. On "could not run" the hook emits a diagnostic to stderr naming which check failed and why, then exits 0. The issue confirms this is safe — it is a PreToolUse hook, so stderr does not disturb the JSON protocol — and ranks it the smallest sufficient fix. The gate still fails open, but it can no longer do so silently, which is the whole complaint: a validator that disables itself quietly costs more than one that is absent, because it is trusted. Both controls are unchanged and pinned by tests: a conforming commit still passes silently, a non-conforming one still exits 2 with its existing block payload. The defect test asserts stderr is non-empty and names the failure; it fails against the pre-fix hook. Verification runs on the remote runner. Refs #3911, #3838 * docs(#3911): document the hook crash-policy contract Reference and Explanation via a new docs/features fragment (FEATURES.md is generated from it), INVENTORY rows for the three new hooks/lib files, and an ARCHITECTURE note on the hooks section. How-To: docs/how-to/declare-a-hook-crash-policy.md, indexed from docs/README.md — a hook author now has to choose and declare a crash policy, which is more than one step and crosses into which harness protocol their hook speaks. It covers allow/deny/crash, writing an ON_CRASH reason that is actually useful, when a deny needs a distinct stderr payload, the two hooks whose harness reads a JSON-body decision and must NOT use deny(), and what to do when a check cannot run at all — with #3838 as the worked example. Refs #3911 * test(#3911): prove the seam actually ships, and stop hand-rolling temp cleanup Two review findings. The acceptance criterion 'hooks/dist/** stays in parity via the build seam (lint:hooks-runtime-build-seam)' was misstated and unmet: that lint checks something else — that a hook requiring a compiled gsd-core/bin/lib module also calls ensureRuntimeBuild(). Nothing exercised that the three new hooks/lib files reach hooks/dist/lib at all. That gap is not theoretical: #770 is a recorded ship-blocking bug where a new hook never shipped because a copy list missed it. The suite now builds dist through the repo's own ensureBuiltHooks(), byte-compares each shipped copy against its source, and spawns a child that requires the SHIPPED dist copy and denies — which is what catches a copy that exists but cannot resolve its sibling registry. gsd-validate-commit.sh hand-duplicated mktemp/run/rm three times; one idempotent trap on EXIT replaces them, guarded so cleanup cannot alter the exit status. Behavior-neutral across five cases, with temp-file counts taken before and after each run. Refs #3911 * fix(#3911): stage transitive hook lib requires, not just one level The remote run returned 7 failures across 3 real causes. The important one is a PRODUCTION bug this phase exposed rather than caused. `writeCursorHooksJson` scanned each hook script for `./lib/X` requires exactly one level deep and never re-scanned the lib files it staged for their own sibling requires. Nothing had a transitive lib dependency before, so the gap was invisible. Adding hook-exit.js -> cli-exit.js -> exit-code-registry.js made real Cursor installs ship a bundle that dies at require time with MODULE_NOT_FOUND. It now walks to a fixed point, and a real installed Cursor hook runs to completion. The staging harness in shared-hooks-dir-resolution hand-copied its fixture, so the injection scanner crashed at require time and its exit-1 was being read as a policy decision. Migrated to copyScriptWithDeps, which walks the require graph — the repo's recorded rule for this class, since adding another copyFileSync keeps it alive for the next person. The missing-lib-source test in cursor-hook-workspace-roots hardcoded which lib file it expected to be named in the abort message; the same throw now fires for a different file first. Its assertion is unchanged in substance — staging still must abort rather than ship a broken hook — only the name is no longer pinned. The last one was my own test asserting an uppercase reason code. Measured against origin/next: the pre-change hook emits the same lowercase 'config_unreadable', so the test was wrong, not the migration. Corrected to the real value rather than making the code match the test. Verification runs on the remote runner. Refs #3911 * chore(#3911): regenerate the cursor install-tree golden The staging fix means a Cursor install now correctly carries the two transitive lib files it was silently missing. Additive only — no path was removed. The golden diff is the evidence the packaging defect was real. Refs #3911 * chore(#3911): backfill the changeset PR number Refs #3911 * fix(#3911): a git probe that timed out is not a negative A macOS CI lane failed three deny cases at 2084ms, 2112ms and 2177ms — just past the 2000ms budget these hooks give their git probes. The three that passed took 72ms, 595ms and 651ms. Under shard contention `git rev-parse` overruns, the hook reads the non-zero result as "not a git repo", and allows with exit 0 and empty stdout AND empty stderr. Under load, the guards silently stop guarding. That is ADR-3889's thesis exactly, sitting inside the security hooks this phase is about. The repo had already recognized the class in one place — gsd-cursor-subagent-start.js fail-closed-denies on `git_timed_out` (#3045) — but nowhere else. `hooks/lib/git-probe.js` classifies a probe's outcome, distinguishing a real non-zero exit from ETIMEDOUT, a signal kill, and a spawn failure, rather than folding all four into `status !== 0`. Three guards route their eight git probes through it. The resolution is the same shape #3838 took, and the same one that issue endorsed as smallest-sufficient: fail open, but loudly. **No exit code changes on any path** — a developer on a loaded machine is still not blocked, which keeps #3911's declaration-pass contract intact for exit codes. What changes is that the hook now says on stderr which probe could not answer, instead of presenting silence as a clean verdict. Scope was checked across every hooks/*.js, not just the three that failed: gsd-agent-isolation-guard spawns no git; gsd-statusline's two probes gate only a cosmetic display segment, not an allow/deny decision, and are left alone. The C2 deny assertion was a real-race test — it demanded exit 2 while a slow git legitimately yields 0. It now requires the hook to either deny, or allow with a diagnostic naming the probe that could not run; a silent allow still fails, so the assertion is not vacuous. A deterministic regression stubs git on PATH to sleep past the budget rather than waiting for load to reproduce it. Verification runs on the remote runner. Refs #3911 * test(#3911): a PATH shim cannot intercept the hooks' git spawn on Windows The deterministic timeout regression stubbed git on PATH and asserted the guard reports rather than silently allows. It passes on Linux and macOS and failed on Windows in 83ms and 176ms — the stub was never invoked at all. Mechanism: the hooks call spawnSync('git', args) with no shell:true, so on Windows CreateProcess resolves git.exe only and never a PATH .cmd shim. The git.cmd branch could not have worked and is removed rather than left implying a Windows path that does. Adding shell:true to the hooks to serve a test would change product behavior and widen an injection surface, so the case is skipped on win32 only, with the mechanism written into the skip reason so a future reader does not 'fix' it that way. Linux and macOS keep the coverage, and macOS is where the underlying fail-open was actually caught. Refs #3911 --------- Co-authored-by: sim <sim@local>	2026-08-27 22:21:10 -04:00
Tom Boucher	03b7125293	enhance(#3909): a probe that could not run no longer asserts a verdict (#3944) ... * test(#3909): failing-first suite for the fabricated probe fallbacks Binds the four fabrication sites found by executing the surfaces (ADR-3889 failure class (c)), each with a positive control so an over-firing fix goes red: - the blocking api-coverage.verify-pre gate certifying "no external-API integration" from a zero-byte phase scope - the assumption-delta query route scanning an unresolvable phase section as the empty string and reporting it as an examined negative - both capability fragments' probe fallbacks, which append a fabricated verdict rather than replacing, and fire on the legitimate exit-1 negative Verification runs on the remote runner. Refs #3909 * enhance(#3909): a probe that could not run no longer asserts a verdict ADR-3889 Phase 5. Four sites turned a failed or unexamined probe into a confident negative; each now reports what it could not establish. - check api-coverage.verify-pre: a phase with no plan body and no roadmap section ran detection over zero bytes and PASSED the blocking seal gate, certifying "no external-API integration" from input it never read. It now holds with scope_unavailable. The discriminator is bytes examined, never signals found, so a phase whose plans are real and simply carry no API vocabulary passes exactly as before. - query assumption-delta scan: an unresolvable phase section was scanned as the empty string and reported as an examined negative. It now returns {skipped, reason: phase_unresolved}, still at exit 0 — an ADR-2980 degraded result in the payload, leaving the gsd-tools exit projection to P8. - both capability fragments: `|| echo '{"detected":false}'` appended rather than replaced, and fired on the legitimate exit-1 negative, so a correct answer and an honest skip both arrived as two concatenated objects. They now keep the probe's own payload and manufacture only an explicit probe_unavailable skip when the probe produced nothing at all. Every registered outcome is more restrictive on a blocking gate, so this can turn a false green red and never a red green. Docs: FEATURES 156, CONFIGURATION (both keys), references/api-coverage.md seal-time outcome table, and a new how-to for the reason-code vocabulary. Verification runs on the remote runner. Closes #3909 * test(#3909): correct the stale unknown-phase assertion `unknown phase → detected:false, no throw (graceful)` scanned phase 999 against a two-phase roadmap and asserted `detected === false`. That pinned the fabrication as intended behavior: the phase does not exist, so the detector was handed the empty string and its "no core assumption changed" answer described nothing that was ever read. It now asserts the skipped-with-reason shape. The graceful-degradation contract the test was actually protecting — the query succeeds and does not throw on an unknown phase — is unchanged. Found by code review, not by the author. Refs #3909 * docs(#3909): author the FEATURES entry in its generator source `docs/FEATURES.md` is generated by `scripts/gen-features.cjs` from the per-feature fragments in `docs/features/`. The API-coverage entry was edited in the generated file, so the next regeneration silently dropped it. The text now lives in `docs/features/api-coverage-gate.md` and `docs/FEATURES.md` is regenerated from it, leaving the shipped file byte-identical and its content actually derivable. Caught by `lint:generated-sync`. Refs #3909 * test(#3909): bind the skip to "not found", and pin the discriminator The first verification run went red on one case, and the case was wrong rather than the code. `getRoadmapPhaseWithFallback` returns `null` for an unknown phase and for a missing ROADMAP.md, but for a section whose body is whitespace-only it returns the heading line alone — which is not empty. So a body-less section WAS found, and reporting `detected:false` over its heading is a real negative, not a fabrication. The test had assumed the resolver yielded `''` there. Correcting the test rather than the resolver keeps `skipped` bound to the distinction the issue asks for — found versus not found — and avoids diverging `assumption-delta scan` from `roadmap.get-phase`, which the fragment documents as sharing one resolver. Also adds the seeded property the test matrix had promised: for any plan body, the scope read back is whitespace-only exactly when the body was. That pins the gate's discriminator to bytes examined, so it cannot quietly become "no signals found", across unicode whitespace and CRLF. `docs/INVENTORY.md` picks up the reference doc's new seal-time outcome table — surfaced by the co-change gate, not by a lint failure. Refs #3909 * chore(#3909): backfill the changeset PR number Refs #3909 --------- Co-authored-by: sim <sim@local>	2026-08-27 15:50:12 -04:00
Tom Boucher	9410f7e6e6	enhance(#3897): ADR-3473 §8.3 rungs 2-4 — runtime marker, derived Codex sandbox, short-form depends_on (#3941) ... * test(#3897): failing-first coverage for §8.3 rungs 2-4 ADR-3473 §8.3 has four rungs; #3883/PR #3896 shipped the first. This pins the other three RED before any fix. Rung 2 — the install marker has four readers and resolveRuntime is not one. resolveRuntime resolves GSD_RUNTIME > config.runtime > 'claude' and reads no marker at all, while bin/install.js writes one (#2297) and FOUR hand-rolled readInstallRuntimeMarker copies exist: src/model-resolver.cts:65 (cached, with test seams), hooks/gsd-agent-isolation-guard.js:112, and TWICE in hooks/gsd-cursor-subagent-start.js at :346 and :355. Four copies of one rule. Fixtures and seam names mined from PR #3382 rather than re-derived; it implemented this rung and was closed "not on the merits". Rung 3 — the sandbox map, and the fallback that was the real defect. Measured across all 35 files in agents/, deriving workspace-write iff tools: declares Write or Edit: - all 11 CODEX_AGENT_SANDBOX entries derive to their mapped value exactly, zero disagreements — the map carries nothing the contract does not - 24 roles fall through `|| 'read-only'`, of which 16 declare Write or Edit So the map is redundant and the silent fallback is the defect. The maintainer chose to derive but hold those 16 at read-only pending the question of whether Codex enforces sandbox_mode or merely advises; HALT.md records it. T20 asserts the emitted sandbox_mode PER ROLE against a captured baseline, not in aggregate — an aggregate passes while one role silently widens, which is the proxy-instead-of-identity shape this repo names. T24 and T25 fail on a stale hold, so the hold list cannot rot into the subset map being deleted. Rung 4 — shortFormToId, recovered rather than invented. I nearly reported this as another wrong §8.3 claim: `git log -S shortFormToId` returns only documentation commits. That was the wrong instrument. Direct inspection of sdk/src/query/phase.ts at 11918dcc3^ shows five occurrences, and the tests match that code rather than a guess at its semantics — including first-write-wins on a duplicate short form. T43 asserts at the consumer's output: the emitted `waves` map from the real CLI, which pre-fix collapses to {"1":[...]} because every short-form edge is dropped. A unit assertion on resolveDependencyId would have passed throughout this defect's life. Observed RED, this tree: rung 2 11/11 fail — no marker rung, no seams rung 3 T23,T24,T25,T26,T30 fail; T28 fails (validate agents passes a TOML whose sandbox_mode disagrees — it checks presence only) rung 4 T42,T44 fail; T43,T49 fail with waves collapsed to a single wave 1 Green and staying green: T20/T21/T22/T27 as captured baselines, #3885's unresolvable-token warning and wave-verdict suppression, and #3785's display-mapping passthrough. If the third tier over-reaches, those go red — that is their job. Disclosed weakness: T45 (a canonical id with no dash is not short-form indexed) cannot be isolated behaviorally, because planMap always masks it. It is a non-crash boundary pin, weaker than the other rows, and is recorded as such rather than presented as equivalent. Design: .gsd/phase/feat-3897-adr3473-83-rungs/40-design.md Test matrix: .gsd/phase/feat-3897-adr3473-83-rungs/50-test-matrix.md Decision: .gsd/phase/feat-3897-adr3473-83-rungs/HALT.md Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * enhance(#3897): §8.3 rungs 2-4 — one marker reader, a derived sandbox, the third depends_on tier ADR-3473 §8.3 has four rungs. #3883/PR #3896 shipped the first. These are the other three. Rung 2 — the install marker had four readers, and resolveRuntime was not one. resolveRuntime resolved GSD_RUNTIME > config.runtime > 'claude' and read no marker, while bin/install.js writes one (#2297) and four hand-rolled readInstallRuntimeMarker copies existed: src/model-resolver.cts (cached, with seams), hooks/gsd-agent-isolation-guard.js, and twice in hooks/gsd-cursor-subagent-start.js. model-resolver's was already the house idiom, so it was promoted rather than replaced: src/runtime-slash.cts now owns it, and model-resolver plus both hooks delegate. The hooks reach it through ensureRuntimeBuild(), the seam lint-hooks-runtime-build-seam enforces. No import cycle existed - checked both directions before moving anything. The marker is the THIRD rung: env > project config > marker > 'claude'. N1 was checked rather than assumed, and my first reading of it was wrong. A marker holding an unknown name comes back essentially verbatim, which looked like a validation gap. Measured against the env rung with the same inputs - including "../../etc/passwd" and "claude;rm -rf /" - the two are identical, because they share resolveRuntimeNameFromCandidates. N1 asks for exactly that, and it is met. The residual (the shared normalizer normalizes shape, it does not validate against the known-runtime set) is pre-existing on the env rung and plausibly deliberate, since a new runtime should not need a code change. The marker also does not widen the trust boundary in any real sense: it lives inside the install tree beside the code, so anyone who can write it can write runtime-slash.cjs itself. Rung 3 — the map was redundant; the silent fallback was the defect. Measured across all 35 files in agents/, deriving workspace-write iff tools: declares Write or Edit: all 11 CODEX_AGENT_SANDBOX entries derive to their mapped value exactly, zero disagreements. The map carried nothing the contract did not already have, so it is DELETED rather than reconciled. What was actually broken is `|| 'read-only'`, which silently under-granted 24 of 35 roles. 16 of those 24 declare Write or Edit and would widen under derivation. Per the maintainer's decision (HALT.md), they are held at read-only pending the question of whether Codex enforces sandbox_mode or merely advises. Emitted TOML is therefore byte-identical for all 35 roles - asserted per role, not in aggregate, because an aggregate passes while one role silently widens. The hold list self-invalidates. A hold whose role no longer derives broader fails, and so does a hold naming a role with no agents/<name>.md. Without that it would rot into exactly the hand-maintained subset map being deleted, and this commit's own ledger claim would become false over time. Both cases were proved by injecting them and watching them throw. Two committed tests asserted the deleted map's existence and contents. They were pinning the thing being removed, so the tests moved rather than the production code: the 11 role-value pairs survive as a test-local PRE_3897_CODEX_AGENT_SANDBOX baseline, and the assertions now drive the real derivation against real agents/*.md. The coverage is preserved; only its source moved out of production code. validate agents gains checkCodexSandboxPosture, mirroring the existing checkCodexModelPosture: each installed TOML's sandbox_mode must equal the role's expected value, failing with role, expected and found. It previously checked file presence and manifest completeness only, so a TOML whose sandbox_mode disagreed passed. Rung 4 — shortFormToId, recovered rather than invented. I nearly reported this as another wrong §8.3 claim: git log -S returns only documentation commits. Wrong instrument. sdk/src/query/phase.ts at 11918dcc3^ carries five occurrences, and the implementation here matches that code rather than a guess at its semantics - including first-write-wins on a duplicate short form, deterministic from the sorted plan order. It resolves the bare plan number: depends_on: ["01"] now reaches 26-01-auth-hardening. That is a control-flow change, not a diagnostic one - plans that silently collapsed into a single wave 1 now execute in their declared waves, and execute-phase.md consumes those wave values. In-phase only, by construction: the map is built from this phase's rawPlans, so a same-named short form in another phase does not resolve. #3785's display-mapping passthrough and #3885's unresolvable-token warning and wave-verdict suppression are untouched and stay green. If the third tier had over-reached, those are what would have caught it. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3897): close a fail-open I introduced, and wire the posture check to its command Two blockers from review. Both are mine, and one is a security regression my own change created. 1. A held role could escape its hold by editing its own frontmatter. The Codex install loop set the sandbox identity from the agent's frontmatter `name:` field rather than from its filename, so the hold lookup keyed off a value the file itself declares: deriveCodexSandboxMode('gsd-doc-writer', <real file>) -> read-only deriveCodexSandboxMode('gsd-doc-writer-x', <same file, name: edited>) -> workspace-write deriveCodexSandboxMode('GSD-Doc-Writer', <same file, name: recased>) -> workspace-write What makes this a blocker rather than a nit is the DIRECTION. The deleted CODEX_AGENT_SANDBOX map had the identical lookup-key quirk, but it was an allowlist: an unmatched key fell back to read-only, which is safe. The new scheme derives workspace-write from the tool contract and uses the hold as a subtraction, so the same mismatch fails OPEN. I converted a fail-closed quirk into a fail-open one and did not notice; the isolated reviewer proved it by execution. Neither safety net caught it. validateCodexSandboxHolds only checks that <key>.md exists, never that a file's derived identity matches its key. checkCodexSandboxPosture looks the canonical source up by the installed TOML's filename, finds nothing for a renamed agent, and treats it as a custom non-roster agent — silently no violation. The identity is now the FILENAME STEM, which is what validateCodexSandboxHolds already validates and what an attacker editing frontmatter cannot change without renaming the file — at which point the existing validator catches it. The lookup is case-insensitive so a recase does not slip past either. The frontmatter name still drives the TOML body and filename, unchanged; only the sandbox identity moved. All 35 roster files were checked: name matches filename stem everywhere, so a stricter "they must agree or throw" invariant would have been safe against real content. It is deliberately NOT added — it would abort an install on a tampered file where emitting a correctly-derived read-only TOML is the safer outcome. Recorded as a fork rather than decided silently. 2. checkCodexSandboxPosture was exported and never called. cmdValidateAgents (src/verify.cts) called checkAgentsInstalled and checkCodexModelPosture only; grep for the sandbox check in that file returned nothing. So criterion 3 — "validate agents fails on semantic drift, not only on missing files" — was unmet, and `validate agents` behaved exactly as before. That is ADR-3473 Decision 2's named shape: a declared policy with no executor. It also meant the T28 test asserted at the helper's return value while the COMMAND stayed broken — the ADR-3180 Decision 4(b) failure this epic exists to close, committed by me while enforcing it elsewhere in the same epic. Now wired as an additive `sandbox_posture` field beside `codex_posture`, following the sibling precedent exactly. Drift is report-only, not a non-zero exit, because that is what checkCodexModelPosture does — two sibling posture checks disagreeing about whether a violation is fatal would be its own defect. The choice is recorded in a comment rather than left implicit. A consumer-output test now drives the real CLI and asserts on the emitted JSON, and was shown failing before the wiring and passing after. Also corrected a stale artifact: the design's Known limit L1 still claimed rung 3 was not in this deliverable, written while it was halted and false once the maintainer unblocked it. Verified after both fixes: the three bypass probes all return read-only, the per-role table is 35/35 byte-identical, and both hold self-invalidation cases still throw. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(#3897): the marker rung, the derived sandbox, and the bare plan-number depends_on Reference: the runtime precedence ladder in docs/CLI-TOOLS.md gains the install marker rung; docs/COMMANDS.md documents validate agents' new sandbox_posture field; docs/reference/plan-md.md documents that depends_on accepts the bare plan number. Explanation: a docs/features fragment keyed id 3897, so it cannot collide with a concurrent PR hand-allocating a section number, regenerated into FEATURES.md. ADR-3473 §8.3 gains an ANSWER blockquote in the document's own correction style, recording what was measured and built against the section's 2026-08-26 correction - including the qualification that checkAgentsInstalled itself still checks presence only, and the semantic assertion lives in a sibling wired into validate agents rather than folded into it. No how-to. Both user-visible changes are zero-step: a non-Claude install resolving its own runtime, and plans executing in their declared waves, both happen without the user doing anything. docs/how-to/control-the-reported-host-runtime.md covers a DIFFERENT ladder (resolveReportedRuntime / agent_runtime) that this change does not touch, and was deliberately left alone rather than edited by association. No tutorial - nothing multi-step to walk through. docs/AGENTS.md unchanged: it documents Claude-side tools frontmatter, never Codex sandbox_mode, and the emitted tools contract did not change. The prompt layer documents depends_on only by example, not by schema, so nothing there needed editing - and few-shot-examples/plan-checker.md already showed depends_on: ['01'], which now actually resolves. Translated copies of plan-md.md are untouched; the project treats translations as community-maintained. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3897): move the sandbox derivation out of the installer, off the install path, and off a third parser The full suite came back with 26 failures across four files. Three distinct causes, mapped individually rather than assuming the first explained the rest. A. Requiring bin/install.js printed the GSD banner to stdout and corrupted `validate agents` JSON. Unexpected token '', "[36m ██"... is not valid JSON checkCodexSandboxPosture reached deriveCodexSandboxMode by lazily requiring bin/install.js, whose module load prints the ASCII banner. So the command emitted banner bytes before its JSON and every JSON consumer broke, including ten tests that predate this branch. src/ reaching into bin/ was backwards layering that happened to also be loud. The derivation now lives in src/codex-agent-toml.cts - the existing Codex TOML domain module, no new module and no six-gate ripple - and both bin/install.js and src/agent-install-check.cts import it. One owner, which is §8.3's rule applied to the fix for §8.3. B. The stale-hold throw fired on a legitimate partial source dir, and masked a security assertion. validateCodexSandboxHolds treated "this hold's .md is absent from the install SOURCE dir" as a stale hold and threw. A test fixture, or any partial install source, legitimately contains a couple of agents. Worse, it threw BEFORE the path-escape check, so a test asserting that a `../../evil` frontmatter name is rejected got my unrelated error instead of the traversal rejection it was written for. A fail-closed check of mine was hiding a real security check. The "no stale holds, shrink-only" invariant is a property of the repo's canonical agents/ roster, not of whatever directory an install happens to read. It is off the runtime path and enforced where it belongs, in the tests that already existed for it. A partial source dir now installs cleanly, and the evil-name case throws with its own escapes-configHome message again. C. T8 depended on ambient process.env state. The marker/env parity assertion round-tripped through live process.env. It now compares against resolveExplicitRuntime's already-exported dependency-injection parameter - deterministic and hermetic, same claim. Proven still falsifiable rather than assumed: with the marker rung's normalization temporarily bypassed the two rungs diverge ("codex\n../../etc/passwd" vs "codex-../../etc/passwd") and the assertion fails, then passes again once reverted. One correction folded in along the way. The first version of the move added private _extractFrontmatterAndBody/_extractFrontmatterField helpers to codex-agent-toml.cts - a THIRD copy of frontmatter extraction, where the graph already shows two (bin/install.js:2348, runtime-artifact-conversion.cts:893). Adding a third inside the epic whose thesis is one implementation per rule is not defensible. deriveCodexSandboxMode no longer parses anything: it takes (identity, toolsValue) and each caller supplies the tools value using the extractor it already has. Both helpers are deleted. The identity argument is still the filename stem, so the fail-open fix is untouched. Verified after all three: `validate agents --raw` emits parseable JSON with no banner and both posture fields; the four hold-bypass probes still return read-only; the per-role table is 35/35 byte-identical at 26 read-only / 9 workspace-write; the hold list is still 16. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3897): drop a dev-only transitive dep, make the derivation total, retire a stale fallback test Suite down to 7 failures from 26. Three more causes, mapped individually. A. My extractor import dragged in a script that does not exist in an installed tree. Cannot find module '../../../scripts/fix-slash-commands.cjs' Chain: src/agent-install-check.cts imported runtime-artifact-conversion.cjs, which requires command-roster.cjs, whose line 36 requires ../../../scripts/fix-slash-commands.cjs. That path exists in the repo and not in an install, so every test exercising a synthetic install dir died at module load. I picked that extractor for convenience without checking what it pulls in - the same mistake that produced the banner bug, one layer further out. agent-install-check now uses a single-purpose extractToolsLine on codex-agent-toml.cts. That is deliberately NOT a general frontmatter parser: we deleted those helpers a commit ago for good reason, and this reads one line. Verified from outside the repo root that requiring either module prints nothing and does not throw. B. A test pinned the deleted name-based fallback. 'defaults unknown agents to read-only' called generateCodexAgentToml with a fixture declaring tools: Read, Write, Edit. Under derivation an unknown agent with a writing contract correctly derives workspace-write - design row S6, a new writing role gets the contract, not the pin. The behavior it asserted was the silent fallback this rung deleted; identity no longer decides the sandbox. Replaced with two rows rather than a flipped string: no tools declared -> read-only (absence is not a grant), and Write/Edit declared -> workspace-write. Strictly more coverage than the row it replaces. C. The stale-hold check still threw per derivation call. Last commit took the roster-existence check off the install path, but deriveCodexSandboxMode itself still threw when a hold's role did not derive broader FOR THE CONTENT IT WAS HANDED - so it fired on any synthetic fixture for a held role. The throw is gone, and it cost nothing: if a held role's content does not derive broader, the hold pins read-only and derivation returns read-only anyway, so the hold is a no-op and there is nothing to fail about. The staleness invariant is a property of the real agents/ roster, and validateCodexSandboxHolds still enforces it there - confirmed against the real roster after the change, not assumed. deriveCodexSandboxMode is now total: every (identity, toolsValue) including undefined and null returns read-only or workspace-write, never throws. Verified: validate agents emits parseable JSON; the four hold-bypass probes return read-only; the per-role table is 35/35 at 26 read-only / 9 workspace-write. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(#3897): put the rung-3 decision in the shipped docs instead of pointing at an ignored path The ADR entry and the feature fragment both ended their rung-3 explanation with "see .gsd/phase/feat-3897-adr3473-83-rungs/45-decision-rung3-sandbox.md". That directory is gitignored (.gitignore:55), so the rationale for holding 16 roles at read-only was reachable only from the machine that produced it. A reader of the ADR got a pointer to nothing. Both now carry the reasoning inline: the criterion asks both that the sandbox derive from the declared tool contract and that no role gain a broader sandbox, and those cannot both hold, because a faithful derivation widens 16 roles the deleted map never listed and that fell through its silent read-only default. The resolution is derive-and-hold - the derivation owns the rule now, each hold is released as its enforcement question is answered, and a hold is reversible where a widened sandbox that turns out to be enforced is not. Checked before assuming this was a defect class: CONTEXT.md cites .gsd/phase/<slug>/40-design.md as its standard Design: provenance line in eight module entries, and four other shipped docs do the same. Citing a phase artifact is an established convention here, so those are left alone. What was wrong was specific to these two: they put load-bearing rationale behind the pointer instead of provenance. docs/FEATURES.md regenerated from the fragment via scripts/gen-features.cjs rather than hand-edited. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3897): close a fail-open, stop a silent mis-resolution, and read a declaration as a declaration Two orthogonal reviews on the shipped sha. Three of the findings are the same failure class this epic exists to close, committed inside it. 1. BLOCKER - the sandbox was decided for one identity and applied to another. bin/install.js derived sandbox_mode for the filename stem and then wrote the result to `${name}.toml`, where name comes from the file's own frontmatter. Make the two disagree and a HELD role's artifact goes wide: rename gsd-doc-writer.md -> gsd-doc-writer-v2.md, keep name: gsd-doc-writer -> stem is unheld, derives workspace-write, lands on gsd-doc-writer.toml add any gsd-*.md whose frontmatter name: is a held role -> clobbers that role's toml with workspace-write Both emit read-only on origin/next, because the deleted map was an allowlist and a miss fell back safe. This is a regression my change introduced. The previous review round moved the HOLD KEY off frontmatter to the filename stem and left the OUTPUT PATH on frontmatter; my own comment at install.js:6985 calls that value attacker-editable, four lines above the line that uses it as the filename. The decision is now made over BOTH candidate identities, most-restrictive wins: if either the stem or the emitted name is held, the mode is read-only. 2. MAJOR - hold matching was toLowerCase() only, so confusables escaped. Turkish dotted/dotless i, fullwidth, NFD, trailing space/NBSP/dot/newline, ./ and ../agents/ all slipped the hold and emitted workspace-write. Identities are now basenamed, trimmed of NBSP/zero-width/control characters, NFKC-normalized and lowercased - and anything still carrying a character outside [a-z0-9._-] is treated as suspicious and derives read-only. We do not enumerate confusables; every shipped roster file is ASCII, so refusing to widen on an identity we cannot recognize is fail-closed with no false positives on real content. 3. MAJOR - the short-form depends_on tier mis-resolved SILENTLY. shortFormToId keyed on the last dash-segment of any canonical id with no constraint that it is a plan number, so a phase holding 09-FIX-auth-PLAN.md made depends_on: ["auth"] bind at wave 2 with zero warnings. This is the worst shape in the epic: the unresolvable-token warning fires on a DROPPED token, so a MIS-RESOLVED one is invisible and the tool reports a confident wave assignment built from a wrong edge. A wrong edge is worse than a missing one. The segment must now match /^\d+$/, which is exactly the contract docs/reference/plan-md.md already documents. This tier was recovered verbatim from the retired SDK lineage, which carried the same defect; we are deliberately NOT preserving it bug-for-bug, and the comment says so, so the next reader does not "restore" it. 4. MAJOR - the derivation was reading a declaration as an absence. extractToolsLine read one line, so a YAML list-form tools: block returned only its first item. Two roster files use list form, and gsd-nyquist-auditor declares Write and Edit there - parsed as "- Read", found no write tool, and emitted read-only. Rung 3's headline claim is that sandbox_mode derives from the declared tool contract; that claim was false for 2 of 35 roles and materially wrong for 1. Reading a declaration as an absence is the silent-drop class this epic exists to close. Renamed extractToolsValue and taught it both shapes. gsd-nyquist-auditor now derives workspace-write and joins CODEX_SANDBOX_HOLDS as its 17th entry, per the standing derive-and-hold decision - so emitted TOML stays byte-identical at 26 read-only / 9 workspace-write while the hold list finally records every role that would widen. A previous pass declined this fix because it moved the count; that inverts the priority. Byte-identity is preserved THROUGH the hold, not by leaving a parser broken. Divergence check, because this is where that bug hides: both paths feeding sandbox derivation - install.js's emitter and checkCodexSandboxPosture - now route through the one extractor. The tools readers in runtime-artifact-conversion and install.js's other frontmatter call sites serve Claude-side emission and do not feed sandbox derivation. Also fixed, each real: the posture check's `found` used a naive whole-file regex where its own sibling uses the block-aware scanner, so prose inside developer_instructions produced a false violation; `found` skipped truncatePostureValue and leaked a 300-char value into validate agents output; deriveCodexSandboxMode's absolute never-throws claim was false for an object with a throwing toString; T49 could not falsify cross-phase leakage (its target phase had its own 01, so a globally-scoped map passed too); T20/N6 iterated a hardcoded table and pinned the FIXTURE size, so a 36th agent would be silently unchecked; three tests reimplemented the code they were testing instead of importing it; and T2-T4 deleted GSD_RUNTIME without restoring it. Verified: hold list 17, gsd-nyquist-auditor derives workspace-write unheld and emits read-only held, roster 35/35 at 26/9, depends_on ["auth"] no longer resolves while ["01"] still does, both identity-bypass cases and every confusable vector emit read-only. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(#3897): the hold list is 17, and the reason the 17th was missing The count read 16 because the derivation could not read the declaration it claimed to derive from: the tools reader was single-line, so a YAML list-form tools: block returned only its first item and gsd-nyquist-auditor's declared Write and Edit were read as an absence. Both the ADR entry and the feature fragment now carry the corrected count and the reason for it, rather than a silently updated number. Deriving from a declaration you cannot parse is not deriving, and a flattering count is worse than a wrong one because it looks settled. docs/FEATURES.md regenerated from the fragment. Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#3897): backfill changeset pr number Refs #3897 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-27 15:19:01 -04:00
Tom Boucher	1e67ec9737	enhance(#3908): the scanners distinguish an empty diff from one they could not compute (#3937) ... * feat(#3908): the scanners distinguish an empty diff from one they could not compute collect_files ended 2>/dev/null || true, which destroyed the evidence three ways: the redirect discarded git's diagnostic, the pipe replaced git's status with grep's, and || true forced success regardless. Four distinct conditions - an established-empty diff, a bad ref, no repository, and a repository with no commits - all reported clean, and a secret scanner reporting clean because git failed is indistinguishable from an all-clear to any gate consuming it. git now runs separately from the filter so its status and diagnostic both survive. An established-empty diff exits NO_INPUT; a scope that could not be established exits UNAVAILABLE; the usage sites move off 2 to USAGE. || true is retained on the filter alone, where it is correct: a diff of only images is empty, not failed. Codes are sourced from a generated shell fragment rather than written into three scripts, so a re-allocation cannot desync them, and a missing fragment fails loudly instead of falling back to literals. The security workflow is updated in the same change: without it, a docs-only PR would newly fail the job. * fix(#3908): keep scanner stderr out of the file list, and drop try/finally from test bodies Capturing git and find output with 2>&1 was right for the failure path but wrong for the success path: a warning emitted alongside a successful diff flowed into the file list and was treated as a filename. stderr is now captured separately, forwarded as a warning on success and as the diagnostic on failure, and never folded into the list. Also converts the control tests' try/finally blocks to t.after(), which CONTRIBUTING bans inside a test body because it masks failures. * chore(#3908): backfill changeset pr number * docs(#3908): record the scanners' four-outcome exit contract SECURITY.md is root-level, so the docs gate correctly held: a Changed fragment owes a file under docs/. The contract also belongs where the feature is described, as REQ-SCAN-INJ-05. docs/FEATURES.md is GENERATED from per-feature fragments (#3840) - the first edit went into the generated file and gen-features --check caught it, which is the same edit-the-output drift this epic exists to close. The fragment is the source; FEATURES.md is regenerated. --------- Co-authored-by: sim <sim@local>	2026-08-27 13:11:13 -04:00
Tom Boucher	929e02cb2c	enhance(#3885): no silent swallow, and no verdict manufactured from dropped data (#3925) ... * test(#3885): failing-first coverage for the depth bound and the manufactured wave verdict ADR-3473 §8.5 says a swallowed failure may not become an authoritative-looking answer. Three families do exactly that today; this commit pins each one RED. Measured on this tree, 2026-08-27: intel query, .planning/intel/file-roles.json nested 12000 deep -> exit 1, "Error: Maximum call stack size exceeded" searchJsonEntries / matchesInValue carry no depth parameter at all. The MAX_JSON_SEARCH_DEPTH = 48 bound existed in the retired SDK lineage (sdk/src/query/intel.ts at 11918dcc3^) and the surviving .cts lineage never received it. same fixture nested 48 and 49 deep -> both return total=1 at exit 0, truncated=undefined Nothing distinguishes "searched to the bottom" from "stopped looking". query phase-plan-index, a plan whose depends_on names an unresolvable token -> warnings: ["Plan 03-02: declared wave: 2 but depends_on DAG places it in wave 1"] The token is never mentioned. computeDependencyLevels drops the edge with `if (!resolvedDep) continue;`, every plan becomes a root, and the tool then reports the author's correct wave: as the thing that is wrong. countPhasePlansAndSummaries with fs.readdirSync throwing EACCES -> hasContext:false, indistinguishable from a phase that simply has no CONTEXT.md. context_read_error is undefined. The shapes these tests assert against, chosen here so the implementation has a target rather than inventing one later: `truncated: boolean` on the intel query result, `unresolved: Array<{plan, token}>` from computeDependencyLevels, and `context_read_error: string | null` per analyzed phase. Deliberately green, and they must stay that way — each stops the fix from over-firing: depth 48 is found and NOT flagged truncated (the ceiling is inclusive) a shallow miss reports no truncation (noise control, N1) 10,000 siblings at depth 2 are unaffected (the bound is DEPTH, N2) a genuine wave: mismatch on a fully-resolved DAG still warns (N3) a genuinely missing directory is absent, not an error the emitted depends_on display mapping still passes an unresolved token through verbatim — already pinned by the existing #3785 test, so no duplicate was added T31 asserts at the consumer's output per ADR-3180 Decision 4(b): it runs the real CLI and reads the emitted JSON, because a unit assertion on computeDependencyLevels would have passed throughout #3427's life. Design: .gsd/phase/feat-3885-no-silent-swallow/40-design.md Test matrix: .gsd/phase/feat-3885-no-silent-swallow/50-test-matrix.md Refs #3885 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * enhance(#3885): no silent swallow, and no verdict manufactured from dropped data Implements ADR-3473 §8.5. A failure or a gap in the input stops being absorbed into an output that reads as authoritative. The recursion bound, restored but NOT verbatim (src/intel.cts) MAX_JSON_SEARCH_DEPTH = 48 is threaded through searchJsonEntries and matchesInValue, which carried no depth parameter at all. The bound existed in the retired SDK lineage (sdk/src/query/intel.ts at 11918dcc3^) and the surviving .cts lineage never received it — §8.3's "a consolidation may not delete an invariant along with the surface that held it", demonstrated. Measured before: a .planning/intel file nested 12000 deep exits 1 with "Error: Maximum call stack size exceeded". Reachable from a project document. The original returned a bare `false` at the ceiling. Restoring that verbatim would trade a crash for a silent "no match" when the truth is "I stopped looking" — the same class this epic exists to close, and ADR-3473 Decision 4 forbids it. So the bound carries a truncation signal: nesting 47 -> found, truncated false nesting 48 -> found, truncated false (the ceiling is inclusive) nesting 49 -> not found, truncated TRUE nesting 12000 -> exit 0, truncated TRUE, no RangeError A shallow document that simply has no match reports truncated FALSE — the flag means "I stopped early", never "I found nothing", or it would be noise. The bound is on DEPTH: 10,000 siblings at depth 2 are unaffected. The dropped edge is named, and stops being blamed on the author (src/phase.cts) computeDependencyLevels dropped every unresolvable depends_on token with a bare `continue`. Each drop makes a plan a root, so the whole phase collapses to wave 1 — and cmdPhasePlanIndex then reported the author's CORRECT wave: as the thing that was wrong. Before: warnings: ["Plan 03-02: declared wave: 2 but depends_on DAG places it in wave 1"] After: warnings: ["Plan 03-02: depends_on token \"nonexistent-token-3427\" does not resolve to any plan in this phase — edge dropped, wave placement for this plan may be unreliable"] The suppression is PER PLAN, never blanket: a plan with a fully-resolved DAG and a genuinely wrong wave: still gets the mismatch warning. resolveDependencyId stays two-tier — the shortFormToId third tier is §8.3/Phase 6's rule and is deliberately not built here. The emitted depends_on display mapping still passes an unresolved token through verbatim (#3785). No artifact from failed inputs (gsd-core/workflows/review.md, #3352) A failed lane leaves no result file, so "every lane failed" is exactly "the aggregate JSONL has zero lines" — the gate condition already existed as a byproduct. REVIEWS.md is no longer written in that case, and the commit step is skipped with it. A budget-SKIPPED lane also leaves no file and is NOT counted as a failure. Per-lane output and non-empty .err are preserved to .review-diagnostics/ before `rm -rf "{run_dir}"` destroys the only record that the lanes failed at all; the commit step names one file, never a glob, so the diagnostics are not swept in. Unreadable is not absent (roadmap.cts, gap-checker.cts, init.cts x2) Four callers collapsed an EACCES on a phase directory into [] and reported hasContext:false — byte-identical to a phase that simply has no CONTEXT.md. Each now names the directory it could not read. A genuinely missing directory stays absent rather than becoming an error, which is what keeps the fix from over-firing. Fatal errno folded into a retry set: audited, no defect found Reported as a verified negative rather than padded with a change. withPlanningLock was fixed by #1884/PR #3472; acquireStateLock by #3776; atomicRenameWithRetry and estimate-cli's renameWithRetry are correct by construction — bounded set {EPERM,EBUSY,EACCES}, bounded attempts, and they return or rethrow the final error rather than swallowing it. estimate-cli's sole caller surfaces that rethrow as write_error in its JSON output. Manufacturing a diff to make the checkbox look worked-on is the Goodhart outcome Decision 6 exists to prevent. Disclosed: R46 (the commit step names one file, never a glob) is a real regression guard but is NOT independently failing-first — the commit fence is byte-identical pre- and post-fix, so it only fails pre-fix through its shared extraction dependency. Recorded rather than claimed as fail-first. Design: .gsd/phase/feat-3885-no-silent-swallow/40-design.md Test matrix: .gsd/phase/feat-3885-no-silent-swallow/50-test-matrix.md Refs #3885 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3885): escape untrusted tokens, and stop cleanup destroying unpreserved evidence Two review findings, both real, both in my own change. An isolated adversarial review found the evidence-preservation block never checked mkdir/cp exit status while `rm -rf "{run_dir}"` ran unconditionally in a SEPARATE fenced block. A disk-full or unwritable phase directory therefore still destroyed the only copy of the failed lanes' output — reintroducing the exact #3352 data loss this item exists to stop, inside the fix for it. Preservation and cleanup are now one block, because each fenced block is a separate execution and a shell variable cannot carry between them. mkdir -p and each cp are exit-checked; cleanup runs only when preservation succeeded, and a failure warns naming the intact run directory. "Nothing to preserve" is not a failure and still cleans up. Driven three ways: success removes run_dir, failure leaves it intact with the warning, nothing-to-preserve removes it. The failure is induced by a file-vs-directory conflict rather than chmod 0o000, which root bypasses. The new unresolved-depends_on warning embedded a user-authored token verbatim: warnings: ["Plan 03-02: depends_on token \"evil Plan 03-01: FORGED WARNING\" does not resolve ..."] The JSON wire form is safe, and the security reviewer judged it non-exploitable for that reason. It is escaped anyway through formatDiagnosticToken — the helper #3884 added one phase earlier for exactly this class. warnings[] is an array a consumer naturally prints line by line, and not reusing the sibling fix is the generative-fix-divergence shape this epic exists to close. The same treatment is applied to context_read_error / phase_dir_read_error, which embed a phase directory path a repository can choose, and to the fs error message, which echoes the raw path itself. Known limit L5 recorded: the bound is on DEPTH only. A 300,000-element shallow array yields a 14.5MB reply with truncated:false. Correct per §8.5 and per negative space N2, disclosed rather than left to be discovered. Refs #3885 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3885): unreadable is not absent in intel.cts either, and a corrupt snapshot is not "no snapshot" Blocker from the round-2 isolated review, and it is my own inconsistency: this phase applied "unreadable is not absent" to phase directories and left it broken in the file it was already editing. chmod 000 .planning/intel/file-roles.json gsd-tools intel query <term> -> {"matches":[],"total":0,"truncated":false} exit 0 safeReadJson swallowed every read failure and returned null, so an EACCES was byte-indistinguishable from an absent file AND from a genuine no-match. Now it separates three states: ENOENT stays silently absent, because not every project has every intel file and intelQuery loops over all of them expecting misses; EACCES/EIO and malformed JSON are both surfaced naming the file. A corrupt intel file previously read as "no matches" too — same defect, same fix. Threading that outcome through the other three callers found something worse than the reported case. intelDiff returned no_baseline:true for a corrupt or unreadable snapshot — not a silent failure but an actively FALSE verdict, telling the caller they never took a snapshot when they did. That is §8.5's headline case, so it is fixed and tested rather than noted. intelStatus and intelApiSurface collapsed the same way; intelApiSurface additionally printed a "not yet populated" banner that was simply untrue. Every row is failing-first, including the absent-file ones — the field is new, so it does not exist pre-fix at all. Those rows are not pre-fix pins; they pin that the fix does not OVER-fire on the ordinary absent case, which is what would turn this into noise on every project lacking an intel file. IO failure is injected by monkeypatching fs and restoring in finally, never chmod 0o000 — root bypasses mode bits, so the reviewer's manual chmod repro is not reproducible as a test. Refs #3885 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test(#3885): build the pathological intel fixture as text, not by stringifying a nested object The remote runner came back red on Linux with two failures, both T4: deeplyNestedIntelDoesNotOverflowTheStack, while the same test passed on macOS. The product was never at fault. writeNestedFixture(12000) built a 12,000-deep JavaScript OBJECT and then JSON.stringify'd it. JSON.stringify recurses once per level, so it overflowed the TEST PROCESS's stack — the error was thrown before the CLI was ever spawned. Linux's container stack is smaller than macOS's, which is the whole of the platform difference. Measured, with the same document built as JSON TEXT so nothing in the building process recurses: depth=100 rc=0 truncated=true depth=5000 rc=0 truncated=true depth=12000 rc=0 truncated=true depth=60000 rc=0 truncated=true V8 parses this shape iteratively; only stringify recurses. The bound works at every depth tried. The fixture is now built by string concatenation. That is also the more faithful input — a real deeply nested JSON document on disk is exactly what the bound guards, where a stringified object was only ever a way to produce one. The depth stays 12000. Lowering it would have made the test pass by weakening it to accommodate a fixture bug, and 12000 is a legitimate pathological input the product handles. T4 remains a genuine fail-first: rebuilt against the parent of the commit that added the bound, the string-built depth-12000 fixture still drives the CLI to rc=1 with "Error: Maximum call stack size exceeded". A comment records why the fixture is text, so it is not "simplified" back into a macOS-green / Linux-red test. Refs #3885 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#3885): backfill the changeset PR number Refs #3885 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test(#3885): normalize path separators before splicing into the workflow's bash CI red on one lane — test (windows-latest, 24, shard 3/3). macOS, Linux and the remote runner were all green. AssertionError: commit must name the single REVIEWS.md file; got: --files C:UsersRUNNER~1AppDataLocalTempgsd-3352-phasedir-mOKmuy/03-REVIEWS.md Every backslash in C:\Users\RUNNER~1\AppData\Local\Temp\... was eaten. The harness spliced an OS-native temp path into the extracted bash, and bash consumes \U, \A, \L and \T as escapes on an unquoted expansion. The same loss broke RUN_DIR, so "rm -rf" targeted a path that never existed and the run directory survived — which is the other two assertions. This is a fixture defect, not a product one, and that was checked rather than assumed. In production the phase directory is toPosixPath-normalized at every call site that serializes it (bin/lib/init.cjs:951, 1381, 1461, 1529, 1595), and the run directory is created by "mktemp -d" running inside the bash block itself (gsd-core/workflows/review.md:163), which emits POSIX-style output even under Git-Bash on Windows. Neither ever carries a backslash where the workflow reads it. The file's pre-existing #3034 harness splices raw native paths too, but only ever inside double-quoted assignments, so it never tripped this — my new harness followed that convention faithfully into the one place where it does not hold. Both now splice through toPosixPath from shell-command-projection, the established seam, which is a no-op on POSIX and mirrors what production does. No assertion was weakened. "commit must name the single REVIEWS.md file" and "the run dir must still be destroyed" still assert exactly that; only how the fixture supplies its path changed. Nothing is skipped on Windows — a t.skip() here would have hidden the question of whether the exposure was real, which is the question that mattered. Driven both ways: a synthetic C:\Users\RUNNER~1\... input reproduces the exact CI string when unfixed and yields C:/Users/RUNNER~1/... when fixed; a POSIX input produces a byte-identical shape, proving the normalization is idempotent. Refs #3885 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * test(#3885): stop the harness making the deleted run dir its own cwd Windows shard 3/3 stayed red after the separator fix, on two assertions the separator fix never touched: AssertionError: the run dir must still be destroyed AssertionError: nothing to preserve is not a failure — run dir must still be removed The separators were a real bug and fixing them fixed the --files assertion. They were not this bug, and two CI cycles went into the wrong axis before I stopped converting path forms and looked at what the harness actually does. runWriteReviewsFlow passed cwd: runDir to runHook, so the child bash process's working directory WAS the directory the block under test then removes with rm -rf "$RUN_DIR". POSIX allows a process to delete its own cwd — verified locally, cd "$d"; rm -rf "$d" removes it cleanly — and Windows does not: a live process's working directory cannot be removed. So on Windows the directory survived and both assertions failed, on macOS and Linux it vanished and they passed. Nothing to do with slashes. Harness-only. Production never cd's into the run directory; every reference is by absolute path, and RUN_DIR is created by mktemp -d inside the bash block itself (gsd-core/workflows/review.md:165) rather than injected. review.md is unchanged. Fix: the child now runs with its cwd in an unrelated temp directory that the block under test never deletes. Neither assertion was weakened, and nothing is skipped on Windows — the tests in this file carry no platform guard and run there unconditionally, which is how this surfaced at all. Honest limit: the Windows failure mode cannot be reproduced on macOS, because POSIX permits the very thing Windows refuses. The diagnosis is grounded in that documented divergence and in the fact that only the Windows lane failed, but the green outcome on windows-latest is unverified until CI runs it. Refs #3885 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-27 04:12:47 -04:00
Tom Boucher	e20744eacb	enhance(#3884): failure is a value — strict argv, and --pick that signals absence (#3922) ... * test(#3884): failing-first coverage for strict argv and absence-signalling --pick ADR-3473 §8.4 says failure is a value. Three families currently encode failure as success, and this commit pins each one RED before the fix lands. Measured on this tree, 2026-08-26: gsd-tools generate-slug "test" --pick nonexistent -> empty stdout, exit 0 (#3365) gsd-tools audit-open --pick nonexistent_field -> dumps the entire human-readable audit report, exit 0 gsd-tools generate-slug "Hello World" --raw --pick bogus -> prints "hello-world", another field's value, exit 0 gsd-tools query state.planned-phase 3 (positional, no --phase) -> exit 0; STATE.md's "Phase: 2 of 5 (Widget Support)" is overwritten to "Phase: null - READY TO EXECUTE" and the frontmatter gains a corrupted current_phase_name (#3358) tests/pick-flag.test.cjs:27 previously asserted the #3365 defect as the contract ("returns empty string for missing field", success === true). That assertion is replaced by the required behavior rather than deleted. The new parseNamedArgs block calls the spec-object signature that does not exist yet, so it fails today by construction. The 11 existing behavior-lock tests are left untouched here; they are corrected in the implementation commit. C1/C4 assert at the consumer's output - STATE.md's bytes - per ADR-3180 Decision 4(b). A unit assertion on the parser would have passed throughout this defect's life. Design: .gsd/phase/feat-3884-failure-is-a-value/40-design.md Test matrix: .gsd/phase/feat-3884-failure-is-a-value/50-test-matrix.md Refs #3884 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * enhance(#3884): failure is a value — strict argv, and --pick that signals absence Implements ADR-3473 §8.4. Absence, emptiness and failure stop being interchangeable ways to say "I could not answer". parseNamedArgs (src/command-arg-projection.cts) Takes a spec object with a REQUIRED `positionals: number | 'rest'` and returns the hub's Result shape instead of a bare Record. Declaring the positional arity is what makes #3358's call site unrepresentable rather than merely detectable: an unrecognized flag or a token past the declared boundary is now InvalidArgs, naming the offending token and listing the accepted flags. The legacy positional-array call shape throws a TypeError — an internal invariant violation per ADR-3473 Decision 2, so a stale hand-written .cjs call site fails loudly instead of destructuring undefined off a Result. parseNamedArgsOrExit projects a failure onto the caller's error(); it is a projection over the one parser, not a second parser. Measured before, against a STATE.md with a populated phase-2 block: query state.planned-phase 3 (positional, no --phase) -> exit 0; "Phase: 2 of 5 (Widget Support)" overwritten to "Phase: null - READY TO EXECUTE", frontmatter gains a corrupted current_phase_name After: exit 1, `unexpected positional argument "3"`, STATE.md byte-identical. The flag form is unchanged and still updates STATE.md. --pick <field> (gsd-core/bin/gsd-tools.cjs) extractField returns {found,value}, and the pick block no longer shares one catch between "output was not JSON" and "field was absent". An absent field exits 1 with pick_field_absent, naming the field and the keys that do exist; non-JSON output exits 1 with pick_output_not_json instead of dumping the command's entire output. A field that is PRESENT with value null, '', 0 or false still prints at exit 0 — that is an answer, not a failure, and it is what keeps `--pick count` printing 0 on a fresh project. Measured before: `audit-open --pick nonexistent_field` printed the whole human-readable audit report at exit 0, and `generate-slug X --raw --pick bogus` printed "hello-world" — a different field's value, confidently, at exit 0. ADR-3409 Decision 7 explicitly deferred this contract fix to #3473; this is it. The sub-issue's "returns 0 when the count is zero OR absent" wording is superseded by the ADR rule it implements: zero prints 0, absence exits non-zero. Defaulting absence to 0 would demote "could not answer" to "the answer is zero" — the hazard docs/how-to/resolve-unreachable-guard-findings.md already warns against. Guard ledger (ADR-3473 Decision 6) scripts/lint-unreachable-guard-drift.cjs Detector A is RETIRED. Its premise — that a `--pick ... || echo` arm can never fire — is now false, so the shape it forbade is the correct idiom and keeping it would forbid the fix. Detector B (glob-consuming cat/ls, a nullglob mechanism this change does not touch) is retained in full, as are the shared scanner, the escape-marker parser and the baseline. Net: -1 detector, 0 added. The file is not deleted. Call-site audit 45 prompt-layer --pick invocations, every one a plain X=$(...) assignment — none in an if test, && chain, or a pipeline whose status is consumed, and no shell block in workflows/commands/agents/references sets -e. Of the 13 (command, field) pairs the prompt layer reads, 10 are always present; the 3 sometimes-absent ones each sit behind a prior found/existence check. No ADR-3409-class "field the command never produces" remains. Design: .gsd/phase/feat-3884-failure-is-a-value/40-design.md Test matrix: .gsd/phase/feat-3884-failure-is-a-value/50-test-matrix.md Refs #3884 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3884): escape untrusted tokens in diagnostics, and cover five unpinned rows Two review findings, both fixed here rather than recorded as limits. 1. A newline in an untrusted token forged a second stderr line. Before, plain-text mode: $ gsd-tools query state.planned-phase $'foo\nError: forged second line' Error: unexpected positional argument "foo Error: forged second line" After: Error: unexpected positional argument "foo\nError: forged second line" --json-errors mode was never affected — io.error runs that payload through JSON.stringify. Plain-text mode writes 'Error: ' + message verbatim, and the three new InvalidArgs reasons plus the two new --pick diagnostics all interpolate a token that comes straight from argv. Fixed with ONE shared helper, formatDiagnosticToken (src/io.cts), applied at every interpolation site — not a copy per site. It is deliberately NOT applied inside error() itself: several callers in this tree emit intentional multi-line diagnostics, and escaping newlines there would mangle them. The available-top-level-keys list needed the same treatment for a reason the review did not anticipate: `frontmatter get <file>` reads an ARBITRARY user document and echoes that document's own keys into the diagnostic. Verified reachable — a frontmatter key containing a newline reaches the key list — so formatKeyForDiagnosticList is guarding a live path, not a hypothetical one. Ordinary keys still render plain and unquoted; a fix that merely dropped the key would also have passed a "one line" assertion, so the test pins the escaped key's presence too. 2. Five behavior-table rows were implemented but nothing pinned them: B7 a dotted path that dies partway B9 bracket syntax on a non-array B10 a negative array index, in and out of range B14 a JSON root that is not an object B17 an @file: payload over 50KB B17 is the load-bearing one. output() writes @file:<path> instead of inline JSON past 50000 characters, and --pick resolves that BEFORE parsing; with no test, a future reordering of those two steps turns every large result into a false pick_output_not_json. The fixture seeds 1200 phase directories and measures the payload at 62474 characters, asserting the spill actually happened rather than assuming it. Refs #3884 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3884): correct the strict-argv surface against a full verification run The first full run came back with 90 failures across 12 files, none in the new tests. They were the argv surface telling me what it actually is. Ten root causes; each classified before anything was changed. I over-implemented, and that is reverted. ADR-3473 §8.4 says parseNamedArgs rejects "unrecognized and positional tokens". It says nothing about a value flag whose value is missing. Making that an error was my design decision, not the rule, and it broke a deliberately recorded contract: `--prd` with no value resolving to null (tests/init.test.cjs emptyPrdValueIsFalsyAndTreatedAsAbsent, row B5; tests/section-manifest-init-facts.test.cjs "flag-shaped value"). The "requires a value" branch is deleted outright rather than kept behind an option — an unused strictness mode is speculative generality. Unknown-flag and unexpected-positional rejection, which is what §8.4 actually mandates, is unchanged. --wave needed a third flag kind the original design did not anticipate. `--wave N` is documented (commands/gsd/execute-phase.md:4,48) and the shipped workflow reconstructs and passes it (execute-phase.md:84), while #2932 records token-PRESENCE semantics: the CLI cares only that the flag appeared, and the value belongs to the workflow layer. That is neither a boolean flag nor a value flag, so `optionalValueFlags` now exists — presence-only in `data`, and the validation cursor consumes a following non-flag token so it is not reported as a stray positional. Every other declared boolean flag was checked against every argument-hint and prose usage in commands/, workflows/, agents/ and docs/; `--wave` is the only one of this shape. Five tests were pinning forms that never worked. tests/adr857-core-without-capabilities.test.cjs passed `init plan-phase --phase 01-stub`, but the documented form is positional (docs/CLI-TOOLS.md:776) and the handler reads args[2] — which for that form is the literal string "--phase". Measured on the pre-fix build against a real .planning/phases/01-stub/ directory: init plan-phase 01-stub -> phase_found=true init plan-phase --phase 01-stub -> phase_found=false The test asserted only exit 0 and key presence, so it had been green while proving nothing about phase resolution. Corrected to the documented form and strengthened to assert phase_found === true. Same class in state.test.cjs (`--plan-count`, a flag that does not exist; the real one is `--plans`), milestone-archive.test.cjs (`init new-milestone --json`, silently ignored), and concurrency-safety.test.cjs (a bare positional field name whose OR-assertion passed because a whole-document dump happens to contain the substring it looked for). Six handlers had no argv validation at all — the same #3358 shape this phase exists to close, found while fixing the rest: init verify-work / phase-op / review / todos / remove-workspace read args[2] with nothing checking the rest, and validate health read --repair/--backfill through a bare args.includes() scan that bypassed the parser entirely. All now go through the seam, so the flag has one owner. tests/init-debug.test.cjs rows C4/C5 asserted that an unrecognized flag must NOT fail. That is the behavior §8.4 removes, and Decision 8 says a caller's local expectation does not override §8, so they are inverted and renamed — a test still called "ignores an unrecognized flag" while asserting rejection would be its own defect. Row C6's point is its PWNED canary; that assertion is kept verbatim and only its exit-status expectation changed, because the hostile token is now rejected rather than absorbed. The blast-radius estimate in 40-design.md is corrected rather than quietly left wrong. get_impact reported MEDIUM / 8 symbols upstream, and that was accurate for what the graph can see — parseNamedArgs's callers. It cannot see that those callers' handlers accept argv shapes wider than the code reading args[2] suggests, which is where the real surface was. Refs #3884 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3884): withdraw the validate-health tightening, finish the A2/A3 revert Second full run: 46 failures, down from 90. Four causes, two of them mine. Reverted `validate health` entirely — it was scope creep, and it broke a real flag. ~30 of the 46 read `unknown flag "--json"; accepted: --repair, --backfill`. The previous commit routed `validate health` through the parser on the reasoning that a flag should have one owner. That was wrong twice over: §8.4 names parseNamedArgs and count queries, and `validate health` was never a parseNamedArgs call site — it read its flags, just not through the parser, so it had no silent-drop defect to fix. Tightening it omitted `--json`, which the health-diagnostic suites use heavily. The handler is now byte-for-behaviour back to its pre-branch form. `validate context` stays converted: it genuinely was a call site, and its `--json` is now declared rather than read by a second `args.includes` scan. The five handlers that had NO validation at all — init verify-work / phase-op / review / todos / remove-workspace — stay fixed. Those read args[2] with nothing checking the rest, which is the #3358 shape this phase owns. Finished the A2/A3 revert. Three tests still encoded the deleted "a value flag with a missing value is an error" rule, including one added by the previous commit for that rule. All three now assert the reverted null contract, and the ones whose titles said "rejected" are renamed — a test named for a contract it no longer asserts is its own defect. `--wave=` and `--wave --weird` are correctly rejected. Neither is documented in commands/gsd/execute-phase.md, gsd-core/workflows/execute-phase.md or docs/, and neither is emitted by the shipped prompt layer, so both are unrecognized tokens that §8.4 mandates rejecting. `doesNotConsumeFollowingFlagAsWaveValue` keeps the property it exists for — asserted directly now, at the parser, that `--wave` does not swallow a following flag as its value — and only its exit-status expectation changed. A contradiction inside this branch, surfaced by the audit and resolved the safe way. Two pre-existing #3573 tests call `state begin-phase '2'` and `state planned-phase '2'` with a bare positional, relying on the old permissive parser to ignore it. This branch's own #3358 regression test requires that exact argv to be REJECTED. The two are mutually exclusive. Widening the router to accept a bare positional — mirroring complete-phase — would have silently re-opened #3358, and was verified to do exactly that: with the widened router, `query state.planned-phase 3` returned exit 0 and wrote current_phase_name again. It is reverted. docs/CLI-TOOLS.md:116 and docs/COMMANDS.md:2192 document only the `--phase N` form for both verbs, so the two #3573 tests move to it. Their assertions were never about the call shape — only that total_phases survives the resync — and both still pass. complete-phase is untouched: its bare positional IS documented, and it keeps the dynamic boundary and the negative-space note that record why. The audit that produced this is in the PR body: for every handler whose declaration changed, the flags it reads anywhere in its body, the flags the shipped surface documents, and the shapes the suite passes, compared. The `--json` miss was a pattern, not an accident — declaring a handler's flags from its parseNamedArgs call alone misses whatever it reads elsewhere. Refs #3884 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#3884): backfill the changeset PR number Refs #3884 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-27 00:12:13 -04:00
Tom Boucher	fb2d122d7f	feat(#3841): assert gsd-tools identity on every state-mutating verb (#3848) ... * feat(#3841): assert gsd-tools identity before any state-mutating verb only this package publishes. The path-based branches — a project-local install, a runtime config directory — had no such guarantee; they trusted their configured location. This closes them. Mechanism: once resolution finishes, and before any verb runs, the preamble probes the tool it picked with `runtime-identity --raw` and matches the answer with a shell `case` pattern ANCHORED to the start of the compact payload (`{"packageName":"@opengsd/gsd-core"`). An unanchored substring match accepts the decoy `{"packageName":"get-shit-done-cc","note":"@opengsd/gsd-core"}`, which any colliding package could publish. The outcome is exported as the two-valued `GSD_IDENTITY_STATUS` (`ok`/`unverified`), so the gate is asserted on a VALUE rather than on warning prose. Rollout is warn-then-fail per the #3146 ruling: `unverified` prints one line naming BOTH causes and continues, because `no_identity_verb` cannot tell a foreign package from an `@opengsd/gsd-core` older than the verb, and at rollout the old-version case is the common one. The blocker was byte budget, not design. The preamble is inlined into 112 shipped files and several sat within single-digit bytes of frozen ceilings (`gsd-verifier.md` 16 bytes, `gsd-executor.md` 33, `execute-phase.md` 234); a first attempt broke five of them. What made room was collapsing the resolver's twenty near-identical `elif [ -f … ]` arms into one candidate-list helper (`_gsd_at`), which buys far more than the assertion costs. The preamble is now 2,624 bytes against 4,500 — a net 1,876 bytes SMALLER per inlined file, so every capped file moved away from its ceiling rather than toward it. No cap raised, no size-budget exception added, no override token emitted. Resolution order, every runtime-home probe, the `unset -f gsd_run` re-source fix, the fail-closed `exit 1`, and the `CLAUDE_ENV_FILE` persistence are all preserved byte-for-byte in substring terms; the snippet still begins with `_GSD_SHIM_NAME=` and still ends with `fi`, which the parity extractors anchor on. `gsd-core/references/gsd-run-resolver.md` is re-synced byte-equal. Also fixes two stale claims found in passing: CONTEXT.md and FEATURES.md both described an `[ -x ]` guard as the load-bearing re-source defense. That guard was tried and REMOVED in #3831 — it rejected the bare function name, fell through every branch, and hit `exit 1`, which kills a sourced caller's shell. `unset -f gsd_run` is the actual mechanism. Refs #3841 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3841): pair the anchor's brace by requiring a closed identity payload The matrix went red on `tests/new-project-mvp-prompt.test.cjs` — "new-project.md has unbalanced braces: net depth 2" — plus a knock-on report from its parent `bug #1516` describe, which is the same failure counted once at the child and once at the block. Root cause: that guard (:182-189, mirroring #3784 bd53925f) walks characters and increments on `{`, decrements on `}`, with no awareness of shell quoting. It scans `new-project.md` PLUS every `new-project/steps/*.md`, and both `new-project.md` and `steps/auto-mode-config.md` carry one inlined preamble copy — hence net 2 from a snippet that was off by exactly one. The unpaired brace was the `{` inside the single-quoted `case` pattern of the identity anchor, which is correct shell and invisible to a text scanner. Fix in the snippet, not the guard. The pattern now anchors at BOTH ends: `'{"packageName":"@opengsd/gsd-core"'*'}'`. That balances 51/51 with a brace that does real work rather than a cosmetic pair — a truncated payload whose prefix matches now fails too, where before it verified. Safe for any future additive field: a JSON object's own closing brace is always the last character, whatever type the last value has, which is pinned by two negative-space tests (a nested object and an array-valued last key must both still verify). Cost: +3 bytes, against the 1,873 the resolver fold already gave back. The alternative considered and rejected was dropping the literal `{` for a `?` glob. It balances too, but weakens the anchor from "must be an opening brace" to "must be any one character", and the anchor is the entire point. Two guards added so this cannot recur silently: - runtime-launcher-parity (F0) pins brace balance at the SNIPPET, so the next edit to that pattern fails on the file it broke instead of surfacing three files downstream in a test whose name mentions neither the launcher nor this issue. It also asserts depth never goes negative, since a `}` preceding its `{` nets to zero while being unbalanced at every prefix. - runtime-identity gains behavioral truncated-payload and trailing-garbage fixtures, so the added `}` is proven load-bearing rather than merely present. Verified: snippet 51/51 braces; new-project combined net depth 0; the seven other preamble-bearing files with nonzero depth are unchanged from merged next (their own prose, not the preamble, and not in any guard's scan set); all 112 inlined copies and the resolver reference re-synced byte-equal; sync:launcher idempotent on the second run. Refs #3841 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#3841): backfill changeset PR number Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-25 01:05:53 -04:00
Tom Boucher	36375513b9	feat(#3840): generate docs/FEATURES.md from per-feature fragments (#3845) ... * feat(#3840): generate docs/FEATURES.md from per-feature fragments docs/FEATURES.md was hand-maintained, and every feature PR wrote into two shared mutable cells: the '### N.' heading whose integer was hand-allocated at authoring time, and the hand-maintained table of contents. Concurrent PRs all picked the same next integer, and two PRs adding differently numbered features still collided on the TOC. #3831 was renumbered 165 -> 166 -> 167 -> 168 across successive rebases, each collision also costing a full matrix verification run because the sha-keyed pass marker dies with the rebase. Mechanism: one fragment per feature at docs/features/<slug>.md carrying id/title/group (and an optional order) in frontmatter, consolidated by scripts/gen-features.cjs --write|--check into a marker-delimited region of docs/FEATURES.md that holds BOTH the TOC and every section body. Group headings and their order are derived too - a group sorts by its lowest-ordered member - so there is no shared registry to edit either; optional per-group prose lives in docs/features/_groups/<slug>.md. A contributor adds exactly one new file. Wired into regen:derived and lint:generated-sync alongside the eight existing generators, matching gen-adr-index.cjs's CLI shape and typed-REASON reporting. Migration froze all 168 existing numbers verbatim: identical section set, identical order, identical bodies. Two defects found in the tree are fixed inline rather than carried forward - the '## Related' block had been spliced into the middle of the document, orphaning §142's Reference line, and four inbound anchors were already broken on next (FEATURES.md#runtime-identity in two files, and #143-spec-phase-edge-completeness-probe off by one). Since the repo has no link checker, --check now validates every inbound FEATURES.md#anchor by resolved target, so that class cannot ship silently again; locale FEATURES.md files resolve elsewhere and stay out of scope. Refs #3840 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3840): carry upstream §69 delta into its fragment and harden the generator Review found section 69 missing '[--strict]' and REQ-STATE-05/06 versus origin/next. Root cause was a stale base, not extraction loss: those lines landed in 394bf384b (#3844) AFTER this branch forked at 63abcface, and 'git diff 63abcface origin/next -- docs/FEATURES.md' is exactly that hunk. Merging origin/next auto-applied the hunk into the GENERATED region, which --check immediately reported as stale; the delta is now carried in docs/features/statemd-consistency-gates.md and regenerated from there. --write is now fail-closed. It previously rendered the region even with violations outstanding, warning only on stderr and exiting 0, so a '--write && git commit' chain could commit a FEATURES.md carrying two colliding sections. It now refuses and exits 1; --force is the explicit override and says so in the report. The test that pinned the old behavior now pins the refusal, plus the --force override and its scoping. Marker forgery is rejected at two layers. A fragment body containing '<!-- FEATURES:START' or '<!-- FEATURES:END' is a typed body_forges_region_marker violation (fragments and group notes alike), and spliceIntoFeatures anchors the end boundary with lastIndexOf instead of indexOf, so a marker that reaches the document by any other route can only make the generated region grow, never shrink. Matching is on marker PREFIXES, so a decorated variant comment cannot slip past. Symlinked corpus entries are refused with a typed dirent_not_regular_file rather than read. A fork PR could otherwise commit docs/features/evil.md as a symlink to any readable path and have the generator inline those bytes into the committed docs/FEATURES.md on the next regen. Equivalence re-verified with a method that cannot cancel out. The first check extracted both operands with the same body-normalising helper, so anything that helper dropped was dropped on both sides. The replacement runs two independent passes: a global content-line multiset diff with no per-section logic at all (0 gained, 19 lost, all 19 the stale hand-written mini-TOC links this change deliberately deletes), and a per-section byte-exact body diff carrying a coverage assertion that fails loudly per file when the extractor accounts for fewer lines than the file contains. That assertion caught two blind spots in the checker itself. 168/168 sections present, order identical, one intended body difference (§142 regains the Reference line orphaned by the misplaced '## Related' block). Refs #3840 Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#3840): backfill changeset PR number Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>	2026-08-24 22:49:00 -04:00