enhance(#4154): validate configured entrypoints before reporting install success (#4249)

* test(260903-m7p): expose configured-entrypoint validation gap

* enhance(260903-m7p): validate configured entrypoints before success

* test(260903-m7p): require pre-success entrypoint validation

* enhance(260903-m7p): gate install success on entrypoints

* test(260903-m7p): cover configured entrypoints across runtimes

* enhance(260903-m7p): cover emitted runtime entrypoints

* fix(260903-m7p): sandbox HOME in finishInstall test and fix changeset pr number

- finishInstall(...'cline'...) calls writeNonClaudeDefaults(runtime) in-process
  before the new configured-entrypoint assertion throws. Without a HOME +
  config-location-env sandbox that write resolved through the ambient
  environment and landed in the developer's live ~/.gsd (confirmed absent on
  origin/next baseline, present only on this branch — full-suite HERMETICITY
  WARNING). Sandbox HOME/USERPROFILE and scrub config-location env for the
  duration of the test, matching the existing in-process finishInstall/
  install() pattern in tests/install.test.cjs (#2665).
- .changeset/quick-wasps-sing.md: pr: 0 is a never-backfilled placeholder
  (CONTRIBUTING.md) that fails changeset-lint's invalid_pr check; set to the
  fork PR number until the upstream PR number is known.

* fix(260903-m7p): repair cross-platform and pre-existing shape fallout

- tests/configured-entrypoint-validation.test.cjs: the win32 branch of
  ensureCodexHooksJsonSessionStart writes a .cmd shim under
  <codexRoot>/hooks/; create that dir in the test (the real installer only
  calls this once hooks/gsd-check-update.js already exists) and assert the
  platform-common entrypoint shape instead of a fixed non-Windows array,
  since win32 legitimately emits two entries (cmd shim + script).
- tests/install.test.cjs: finishInstall's shared settings-json return now
  carries configuredEntrypoints/rollbackInstallerMigrations for every
  runtime on that path (trae included, not just Claude/Cursor/Windsurf);
  update the trae install() exact-shape assertion to match.

* fix(260903-m7p): keep .sh interpreter tracking consistent with unresolved bash

configuredEntrypointsForHook's shell branch dropped interpreterCandidates
entirely when resolveBashExecutable returned null, unlike the sibling
portableHooks runner entry a few lines below (which correctly falls back to
the literal 'bash' token). Found via agy adversarial review; verified
unreachable through the current call graph (buildHookCommand's own
resolveBashRunner==null gate already short-circuits before
recordConfiguredHookCommand runs), so this is a defensive consistency fix,
not a live-bug patch — kept for the next caller that does not share that
gate.

* chore(260903-m7p): backfill changeset pr number to the opened upstream PR

.changeset/quick-wasps-sing.md carried the fork PR number (16) as a
placeholder until the upstream PR existed; open-gsd/gsd-core#4249 is now
open, so record its real number per CONTRIBUTING.md's changeset pr-field
convention.

* fix(#4154): track already-registered hooks for entrypoint validation on update

applySettingsJsonHooks registers each guard hook only if absent, so a hook
already present from a prior install keeps its stale on-disk command. The
new entrypoint tracker always records the freshly-computed command for it,
which never matches what is actually persisted, so the exact-string filter
in finishInstall silently dropped it from validation — the Blocker case
this feature exists to catch (an already-installed entrypoint going stale
between installs) was exactly the case it never validated.

Match on the managed script's basename instead, which the persisted
command carries either way, so an already-registered hook stays in the
validated set. Regression test forces this path by mutating a
freshly-installed hook's persisted command before a second install.

* fix(#4154): distinguish an unreadable script from a missing one

validateConfiguredEntrypoints folded an EACCES statSync failure into the
same 'missing' reason as ENOENT, misreporting a real permission problem as
an absent file. Check the error code and report 'unreadable' instead.

* docs(#4154): document entrypoint validation's rollback and PATH scope

CONTEXT.md's Runtime Hooks Surface Module / Installer Module entries had
no mention of ConfiguredEntrypoint/validateConfiguredEntrypoints, despite
bin/install.js x CONTEXT.md being this repo's strongest co-change pairing.

The update-gsd.md how-to overstated what a validation failure undoes: for
Codex/Cursor/Windsurf/Kimi, their own writer already persisted hooks.json/
config.toml inside install() before the aggregate validation call runs, so
there is no rollback path for that write regardless of "where available"
phrasing. Also note that interpreter resolution checks the installer's own
PATH, not necessarily the PATH a hook fires under later (#2979 launchers).

* chore(#4154): point changeset pr field at the fork PR while CI runs there

Mirrors the branch's own prior backfill commit: pr: matches whichever PR
number changeset-lint is currently validating against (fork PR #16 during
the fork-first CI/review loop), flipped back to the upstream PR number
right before the final push to open-gsd/gsd-core.

* fix(#4249): address adversarial-review findings in entrypoint validation

An internal adversarial review (agy/gemini-3.8-flash-high) of the whole PR
found several real gaps beyond the human reviewer's Blocker, verified
against source before fixing:

- Codex's install() result bound rollbackInstallerMigrations to the narrow
  installer-migrations-only rollback instead of restoreCodexSnapshot (#3245),
  the full pre-install snapshot/restore Codex already owns for exactly this
  case — a validation failure discovered outside install() reverted nothing
  of the config.toml/hooks.json that call had already written.
- The register-only-if-absent basename match from the prior fix used a bare
  substring, which an unrelated user command mentioning the same filename
  could false-positive into GSD's validated set — anchored on the
  `/hooks/<basename>` path segment instead.
- nodeCandidates checked raw process.execPath (always true — we're running
  in that process) instead of normalizeNodePath's stable version-manager
  alias, the same one buildNodeRunnerChainToken bakes as its first choice —
  a false green regardless of whether that alias itself still resolves.
- An entry with no interpreterCandidates (Cline's PreToolUse hook, or a
  Windows-Claude .sh hook invoked without a bash runner) runs via its own
  shebang; validateConfiguredEntrypoints checked only file-type, never the
  execute bit. Cline's writer also never reported an entrypoint at all.
- Duplicate (configPath, scriptPath) entries (e.g. Kimi's context-monitor
  hook registered across several events) were validated once per duplicate.

Each fix is covered by a new or extended test; the Codex one required
inlining runCodexInstall's env sandboxing so the rollback closure — which
re-resolves the $HOME-relative skills root live — runs before the sandbox
is torn down, matching how installAllRuntimes' real aggregate gate calls it.

* docs(#4249): document the round-2 entrypoint-validation fixes

Runtime Hooks Surface Module and Installer Module entries now name
ConfiguredEntrypoint's not-executable reason, the normalizeNodePath
alignment, Cline's tracked hook, and which install() result the
finishInstall/installAllRuntimes rollback path actually reverts per
runtime (Codex's full snapshot vs. the others' narrow migrations-only
rollback).

* chore(#4249): point changeset pr field at the upstream PR now that fork CI is green

* fix(#4249): address agy adversarial-review findings

- validateConfiguredEntrypoints: statSync alone never detects a
  chmod-000 script (it only needs parent-dir search permission), so an
  interpreter-invoked entry with an unreadable script passed validation.
  Add an explicit R_OK check for the interpreterCandidates branch only —
  the candidate-less/shebang branch already has its own X_OK gate.
- docs/how-to/update-gsd.md: the blanket "does not revert" claim was
  false for Codex, which reverts config.toml/hooks.json via its full
  pre-install snapshot; qualify it per runtime.
- tests/codex-config.test.cjs: the #4249 rollback regression test
  asserted skills/ and VERSION were reverted but never asserted
  config.toml/hooks.json were too, despite the test's own stated intent.
- CONTEXT.md: qualify which interpreterCandidates entries get
  normalizeNodePath'd (Node hooks only, not .sh/bash) and note Codex's
  Windows .cmd shim as a third candidate-less case that relies on
  extension dispatch, not a shebang.

* fix(#4249): validate Cline's PATH-dependent interpreter, not just its execute bit

Cline's hook is a hybrid: it self-executes via '#!/usr/bin/env node',
so it needs the execute bit (like any shebang-invoked entry), but its
interpreter is looked up on PATH by 'env' at hook-fire time (unlike
every other GSD JS hook, which bakes an absolute node path specifically
to avoid that dependency). The candidate-less/interpreterCandidates
fork treated these as mutually exclusive, so Cline's entry silently
skipped interpreter resolution entirely — a completely missing 'node'
on PATH would still validate successfully.

Add an orthogonal selfExecutable flag so both checks run for entries
that need them. (CodeRabbit finding on the fork rehearsal PR.)

* fix(#4249): address second-round adversarial review findings (opus + agy)

- validateConfiguredEntrypoints: R_OK now runs for every scriptOk entry,
  not just interpreterCandidates ones — a self-executable shebang script
  is still opened and read by its kernel-invoked interpreter, so X_OK
  alone never proved it was readable.
- selfExecutable is now the sole, explicit source of truth for the
  execute-bit check (every producer that needs it sets the flag) instead
  of being partly inferred from an absent interpreterCandidates, which
  Cline's hybrid entry also carries.
- The execute-bit check now skips explicitly on win32 (matching
  resolveExecutableBinary's own carve-out) instead of relying on Node's
  accessSync(X_OK)-as-F_OK no-op, which only protects a real Windows
  machine and not a test that simulates win32 on a POSIX runner.
- bin/install.js: fixed a stale comment claiming no runtime's
  install()-time writes have a rollback path — Codex's does
  (restoreCodexSnapshot) — and added the omitted Cline to both that
  comment and CONTEXT.md's equivalent lists.
- CONTEXT.md: fixed the Cline description left stale by the previous
  commit's selfExecutable addition, and rewrote the validation-mechanism
  paragraph for clarity (writing-for-agents pass).
- docs/how-to/update-gsd.md: split an overloaded 4-clause sentence.
- Removed a fault-injection integration test that could not reliably
  exercise the real installAllRuntimes -> finalize -> rollback wiring
  without fighting the installer's own pre-registration existence
  guards; the constituent pieces remain covered individually.

* fix(#4249): pin platform in X_OK-testing entries so they're deterministic cross-CI-runner

X_OK is a POSIX-only concept, skipped entirely when an entry's platform
is win32 (matching production). Two test entries omitted platform,
defaulting to process.platform — on an actual windows-latest CI runner
that silently skipped the very check they were meant to exercise,
turning 'not-executable' into a false pass. Pin platform: 'linux' so
these are deterministic regardless of which OS runs the suite.

* fix(#4249): classify EPERM the same as EACCES in statSync error handling

Windows raises EPERM (not EACCES) for a parent directory that couldn't
be traversed into — was falling through to 'missing', misreporting a
genuine permission problem as a nonexistent path.

* docs(#4249): address final CodeRabbit doc-completeness findings

- CONTEXT.md: install()'s documented result shape omitted
  configuredEntrypoints; the ConfiguredEntrypoint shape omitted
  selfExecutable.
- docs/how-to/update-gsd.md: the failure-mode sentence omitted
  unreadable and lacks-execute-permission, which the installer also
  rejects.

* fix(#4249): stop double-validating every configured entrypoint on install/update

installAllRuntimes' finalize() already runs assertConfiguredEntrypoints
once over the aggregate set; finishInstall then re-ran the identical
check per runtime in the printSummaries loop right after, so every
entrypoint paid its statSync/accessSync/interpreter-resolution cost
twice on every install and update. Add entrypointsAlreadyValidated to
skip the redundant pass specifically on that path, while leaving the
check intact for any caller that invokes finishInstall directly.

* chore(#4154): point changeset pr field at rehearsal fork PR while CI runs there

* perf(#4249): memoize interpreter candidate resolution across entrypoints

resolveExecutableBinary walked PATH once per (entry, candidate) pair; a
typical install has a dozen-plus entries sharing the same few candidate
lists (process.execPath for JS hooks, bash for shell hooks). Cache by
(platform, candidate) so each distinct pair resolves once per validation
call instead of once per entry.

* chore(#4249): point changeset pr field at the rebased rehearsal fork PR

* fix(#4249): drop entrypoint tracking from the now-dead Codex event writer

#2586 (landed on next after this branch forked) removed install.js's
CODEX_EXTENDED_HOOK_EVENTS registration loop, so ensureCodexHooksJsonEvent
no longer runs during install or update. The ConfiguredEntrypoint records
this branch added inside it were therefore unreachable and untested. Restore
the function to its upstream shape; the entrypoints it used to report were
never collected by any caller.

* refactor(#4249): drop the revalidation bypass flag and the candidate cache

Both were this PR's own micro-optimisations over a set of roughly a dozen
entries. `entrypointsAlreadyValidated` let a caller turn the finishInstall
gate off to save one statSync/accessSync pass; `resolvedCandidateCache`
memoised resolveExecutableBinary across entries that are already deduped by
(configPath, scriptPath). Neither is measurable, and the flag was the only
way to reach finishInstall with validation disabled. finishInstall now
always validates what it is given.

* chore(#4249): point the changeset pr field back at the upstream PR

* refactor(#4249): track settings.json entrypoints without the hooksSurface gate

The install-surface writer only tracked configured entrypoints when the
runtime's descriptor also declared `hooksSurface: 'settings-json'`. Nothing
asserts that axis agrees with `installSurface`, so a descriptor that broke
the coupling would silently pass `configuredEntrypoints: undefined` and drop
that runtime out of the validation this PR adds — reintroducing the exact
'reports Done! over a broken entrypoint' failure #4154 exists to close.

Remove the dependence rather than test it: everything recorded on this path
lands in settings.json by construction, and the registered-command filter
already discards entries no persisted hook references.

* chore(#4249): put the changeset body in the documented two-part format

CONTRIBUTING.md and .changeset/README.md both show
`**<bold change>** — <symptom-led explanation>.`; the fragment was a single
unbolded sentence.

* chore(#4249): point the changeset pr field at the rehearsal fork PR while CI runs there

* fix(#4249): restore the whole manifest-tracked GSD file set on Codex rollback

#3245's snapshot covers config.toml, hooks.json, skills/gsd-*, agents/gsd-*
and gsd-core/VERSION. The install overwrites every other GSD-owned file too —
hooks/, gsd-core/CHANGELOG.md, scripts/, gsd-core/.gsd-runtime, the manifest
itself — before the entrypoint-validation gate runs, so a validation failure
left the new payload sitting on top of the restored old config.

Snapshot the file set the PREVIOUS install's gsd-file-manifest.json claims,
before runInstallerMigrations so the bytes are the true pre-install state, and
restore it from both Codex rollback closures ahead of the per-surface restores.
Files only the failed install introduced are removed, read from the manifest
now on disk. The manifest is already the authoritative record of what GSD owns,
so no second hand-written list can drift out of sync, and user-owned files are
never snapshotted or removed. Every path is confined through
resolveInstallRelativePath, so a hand-edited manifest cannot turn rollback into
an arbitrary-path write.

Non-Codex runtimes are unaffected: the snapshot is gated on the same
tomlConfigInstall + non-minimal condition as #3245's.

* fix(#4249): keep the managed-file snapshot honest in minimal mode and on a bad manifest

Two follow-on defects in the previous commit's snapshot:

- The capture was gated on `!isMinimalMode`, copied from #3245. A core/
  --minimal Codex install still writes gsd-core/, hooks/, scripts/ and the
  manifest, and restoreCodexSnapshot is reachable in that mode (#2695), so the
  snapshot came back empty while the rollback still ran — and its removal pass
  would have deleted every file the new manifest lists. Gate on
  tomlConfigInstall alone, matching where the rollback actually reaches.

- An unreadable or unparseable prior manifest was caught alongside ENOENT and
  treated as a fresh install. That is the same empty-snapshot state, so a failed
  update over a real install with a corrupt manifest could delete its prior
  payload. Track whether the pre-install GSD-owned set is KNOWN: ENOENT means
  known-empty; any other read error or a parse failure means unknown, and the
  restore closure returns without touching anything, degrading to #3245's
  narrower rollback. Deliberately not fatal — a corrupt manifest has to stay
  repairable by reinstalling over it.

Both paths are covered by red-checked regression tests.

* fix(#4249): snapshot Codex skills, agents and VERSION in minimal mode too

commit removed from the manifest snapshot. restoreCodexSnapshot is reachable
for a core/--minimal install (#2695), and its pass-2 sweeps remove every gsd-*
skill dir and gsd-* agent file the snapshot does not claim — so with an empty
minimal-mode snapshot a rollback deleted the whole skills/agents surface with
nothing to restore it from. Codex resolves skills to $HOME/.agents/skills via
the ADR-1239 skills-kind home override, so this is also the reason manifest
`skills/` keys do not resolve under configDir: that surface belongs to this
snapshot, not to the manifest-driven one.

Gate on tomlConfigInstall alone. _codexPreConfigRollback stays null in minimal
mode — doing nothing on an early failure is the non-destructive side.

Covered by a red-checked regression test that plants bytes in an alternate-home
skill file, reinstalls under the core profile marker, and asserts the rollback
restores it.

* fix(#4249): never remove on rollback unless a prior manifest proves what predates the install

Three defects in the manifest-driven Codex rollback, all in its removal half:

- ENOENT marked the snapshot usable, arming the removal pass on a FIRST
  install. GSD may have overwritten a user's file at a manifest-tracked path
  there, and no prior manifest records the difference — so rollback deleted it
  where before it merely left it overwritten. Absent, unreadable and malformed
  manifests now all leave the prior set UNKNOWN and skip removal entirely.

- Membership was tested against the map of files whose pre-install read
  SUCCEEDED, so a tracked file that existed but was unreadable read as
  introduced-by-this-install and was removed. Track the prior manifest's paths
  in their own Set and test against that.

- The unreachable "delete the manifest when there was no prior one" branch is
  gone: usable now implies a parsed prior manifest.

Also adds the end-to-end test the aggregate gate was missing — the four Codex
rollback tests drove the closure directly, proving the restore but not the
wiring. installAllRuntimes(['codex','cline']) under an emptied PATH makes
Cline's `env node` entry fail validation for real, and asserts Codex's payload
comes back.

Test preamble (HOME/USERPROFILE sandbox + config-env scrub) is now one helper
instead of six copies. Both new tests are red-checked.

* test(#4249): use unlinkSync, not rmSync, to drop the manifest in a test

lint:ci's raw-fs.rmSync rule points tests at helpers.cleanup for its
Windows-EBUSY retry budget. That budget is for directory trees; this removes a
single file, which unlinkSync says more precisely and the rule does not flag.

* chore(#4249): point the changeset pr field back at the upstream PR

* fix(#4249): use an unambiguous dedup key and surface partial-restore failures

trek-e's 2026-09-08 adversarial pass flagged two findings in the new
entrypoint-validation/rollback code:
- assertConfiguredEntrypoints' dedup key already used a raw NUL
  separator (introduced in ceebb65f2d), but git/Read render NUL as a
  space, so the key looked like a plain-space join to every reviewer
  that read the diff. Replace it with JSON.stringify([configPath,
  scriptPath]) so the separator is visible and unambiguous.
- restoreManagedFileSnapshot's per-file restore catch block claimed to
  'surface the original error' but only swallowed it, matching (and
  widening) the pre-existing #3245 restoreCodexSnapshot pattern. Add
  an actual console.warn using the existing best-effort-warning
  convention, scoped to just this PR's new function.

* fix(#4249): treat a files-less prior manifest as unknown, not known-empty

agy's gemini-3.8-flash-high adversarial pass (round 5) found and I
reproduced empirically: a structurally-valid manifest missing the
files key (e.g. {"version":1}) parses without throwing, so
Object.keys(undefined || {}) silently read as 'zero files predate
this install' instead of the UNKNOWN state the malformed-manifest
guard exists to produce. Rollback's removal pass then deleted every
GSD-owned file the failed install's own manifest listed, including
ones that predated it — the exact data loss the #4249 CodeRabbit
malformed-manifest fix was supposed to prevent, reachable through a
JSON.parse success instead of a failure. Route the shapeless case
into the same catch-all UNKNOWN path via an explicit shape check.
Regression test reproduces the deletion before the fix and confirms
the file survives after it.

Also extend restoreManagedFileSnapshot's removal-pass rmSync and
final manifest-rewrite catches with the same real console.warn
trek-e's round-4 review asked for on the per-file restore catch —
same rollback function, same operator-facing-signal gap.

* docs(#4249): correct which runtimes actually leave a written config on rollback

agy's completeness audit (round 5, holistic pass) caught this new
paragraph claiming 'for every other runtime, the configuration file(s)
already written during that update are left in place' — false for
Claude Code and other settings.json-based runtimes, whose write never
happens on failure (assertConfiguredEntrypoints runs before
finishInstall's writeSettings). Only Cursor/Windsurf/Kimi/Cline
actually match that description, since they persist their config file
inside install() ahead of the gate. Split the one sentence into the
three actual outcomes; matches the PR body's own accurate Before/After
wording, which this doc addition had drifted from.

* fix(#4249): clean up doc/comment mismatches and dead fields from opus review

Opus critical-code-reviewer + ponytail-review pass on the final diff:
- assertConfiguredEntrypoints carried finishInstall's old docblock
  ("Apply statusline config, then print completion message") from
  before this function was inserted between comment and callee.
  finishInstall already has its own accurate #4249 comment, so the
  stale docblock is removed rather than moved.
- checked: number on ConfiguredEntrypointValidationResult and
  error.configuredEntrypointValidation on the thrown error: the first
  had zero consumers anywhere in the repo, including its own defining
  file, and is removed. The second matches an existing repo
  convention (bin/install.js's installerMigrationRollbackFailures,
  #4249 predates this PR) of attaching structured diagnostic context
  to a re-thrown Error even before a consumer exists, so it's kept.
- finishInstall's own assertConfiguredEntrypoints call is a redundant
  backstop on the real production path (installAllRuntimes's aggregate
  call already validates the superset first), but its comment read as
  though this call alone provided the before-the-write guarantee.
  Clarified rather than removed — it's the only gate for a caller that
  invokes finishInstall directly.

* chore(#4249): split the manifest-driven rollback engine out into #4544

Issue #4154 asked the installer to consume a validation failure "through
the existing rollback mechanism, without a second transaction mechanism".
The manifest-driven rollback widening added during review (capture every
path the prior gsd-file-manifest.json claims, restore those bytes, remove
what only the failed install introduced) is that second mechanism on a
plain reading. It is a real fix for a #3245-era gap, but an independent
one, so it moves to its own bug report and PR.

Removed here:
- bin/install.js: the pre-install managed-file capture block and
  restoreManagedFileSnapshot, plus its call sites in
  _codexPreConfigRollback and restoreCodexSnapshot (99 lines).
- tests/configured-entrypoint-validation.test.cjs: the five tests that
  exercise the manifest engine.
- CONTEXT.md and docs/how-to/update-gsd.md: the sentences describing the
  widened restore. update-gsd.md again documents the #3245 surfaces only.

Kept, because it is #4154's own scope:
- the entrypoint-validation gate itself;
- Codex's install() result binding rollbackInstallerMigrations to
  restoreCodexSnapshot (config.toml, hooks.json, skills/gsd-*,
  agents/gsd-*, gsd-core/VERSION);
- the !isMinimalMode gate removal on that snapshot. Binding the closure
  to the result made it reachable for a core/--minimal install, where its
  pass-2 sweeps delete every gsd-* skill dir and agent file the snapshot
  does not claim; an empty minimal-mode snapshot therefore deleted the
  whole surface with nothing to restore.

The surviving aggregate-failure test now asserts on config.toml, a
surface the #3245 snapshot owns, instead of gsd-core/CHANGELOG.md, which
only the manifest engine restored.

Refs #4544

* test(#4249): cover configured entrypoints through the packed install path

#4154's scope lists install smoke coverage alongside the installer gate —
"assert representative configured entrypoints resolve for supported runtime
profiles". The gate itself (assertConfiguredEntrypoints /
validateConfiguredEntrypoints) is unit-covered by in-process install() calls;
nothing proved the property survives npm pack -> npm install -g -> install.js.

Add Cycle 4 to runSmoke. For each of claude and codex — the two distinct
config surfaces GSD writes launch paths into (settings.json, and hooks.json +
config.toml) — run the tarball-installed installer into a throwaway HOME, then
re-read that runtime's own written config and return the new
ENTRYPOINT_UNRESOLVED code when a script path it names does not resolve to a
file. install-smoke.yml already asserts .code == "ok" on the CLI, so the check
becomes a release gate on every matrix host without workflow changes.

The scan re-derives paths from the written config instead of reusing the
installer's own entrypoint list, and test I shows why that matters: a
registration the installer never touched during a run is invisible to the
in-process gate, so the install exits 0 and only reading the config back off
disk catches the dangling launch path.

* ci(#4249): pack a publish-shaped tarball in the install smoke lane

`npm pack` runs prepack/prepare (build:lib); only prepublishOnly runs
build:hooks. hooks/dist is gitignored, so the tarball install-smoke.yml packs
after `npm ci` carries no hook scripts at all — the lane has been smoking a
package that differs from the published one in exactly the artifacts the
lifecycle smoke is supposed to launch.

That went unnoticed because the lane's init runs `--local`, which registers no
statusline and therefore registers no hook whose target is missing. A
`--global` install on the same tarball exits 1 on #4249's own gate
(`gsd-statusline.js (missing)`), which is what the new configured-entrypoint
cycle performs, so without this step the cycle would report INIT_FAILED
instead of checking anything.

Build hooks before packing so the smoked tarball matches prepublishOnly. The
CLI now reports 16 configured entrypoints for claude and 1 for codex instead
of zero.

* fix(#4249): scope Codex's full snapshot restore to entrypoint failures

Binding Codex's result to `restoreCodexSnapshot` made ANY finalize-stage
exception un-install a Codex install that had already succeeded and already
printed its own "Done!" summary — `rollbackFinalizedInstallerMigrations` wraps
the whole `finalize()` body, not just the aggregate `assertConfiguredEntrypoints`
call.

Nothing documents that. `docs/installer-migrations.md#phase-4-installupdate-integration`
scopes finalize-stage rollback to installer *migrations* ("the executor uses the
journal to restore modified paths"), and this PR's own operator-facing paragraph
in `docs/how-to/update-gsd.md` scopes the Codex config.toml/hooks.json/skills/
agents/VERSION revert to entrypoint-validation failures specifically ("If a
script is missing, unreadable, ... For Codex, this reverts ..."). The wide
behaviour is also incoherent as a transaction abort: the same doc says Cursor,
Windsurf, Kimi and Cline keep the config they wrote inside install().

Concretely: `installAllRuntimes(['codex', 'kilo'])` where Kilo's finishInstall
hits EACCES writing kilo.json rolled Codex's config.toml back to its pre-install
bytes — on an update, silently downgrading a working Codex install to the
previous version while the user had just been told it was Done.

Select the rollback by error kind instead. `assertConfiguredEntrypoints` already
tags its error with `configuredEntrypointValidation`, so the full snapshot
restore runs for that error (and anything downstream of it, including
finishInstall's per-runtime backstop) and the installer-migrations-only closure
runs for everything else. The codex result now also exposes that narrow closure
as `rollbackInstallerMigrationsOnly`; `rollbackInstallerMigrations` keeps meaning
the full restore, so the direct-call contract asserted by
tests/codex-config.test.cjs is unchanged.

Adds a regression test that installs codex+kilo together, injects EACCES on the
Kilo permission write by monkeypatching node:fs (restored in a finally — never
chmod 0o000, which root bypasses in CI), and asserts Codex's config.toml keeps
the bytes the successful install wrote. Verified red against the pre-fix
unconditional path.

Cline cannot host this test: its plan is writesSharedSettings:false +
finishPermissionWriter:null, so its finishInstall performs no write and has no
non-entrypoint failure path. Kilo's configureKiloPermissions runs unconditionally
(unlike OpenCode's, it is not GSD_TEST_MODE-gated) and ends in an unguarded
fs.writeFileSync.

* docs(#4249): sync CONTEXT.md's rollback description with the round-6 narrowing

CONTEXT.md still described Codex's rollback as an unconditional bind
to restoreCodexSnapshot after ff13adc00 scoped it to entrypoint-
validation failures via rollbackInstallerMigrationsOnly and the
configuredEntrypointValidation error tag. Caught during the round-6
PR body pass.

* fix(#4249): stop rollbackInstallerMigrations meaning its own opposite

Codex's install() result bound `rollbackInstallerMigrations` to
restoreCodexSnapshot (the FULL pre-install snapshot restore) and put the
actual installer-migrations-only closure behind
`rollbackInstallerMigrationsOnly` — so for one runtime the unsuffixed name
meant the opposite of what it says, and CONTEXT.md had to concede as much
in prose. Invert it: `rollbackInstallerMigrations` is the narrow closure
for every runtime, matching both its name and the meaning it already has
on next, and the snapshot restore gets its own Codex-only field,
`rollbackPreInstallSnapshot`. The selection in
rollbackFinalizedInstallerMigrations collapses to one line and no longer
needs a fallback chain.

Also in this commit, all against the same rollback path:

- Correct the rollbackFinalizedInstallerMigrations comment. It read as if
  the round-6 narrowing prevented any sibling-triggered revert of a Codex
  install the user has already seen "Done!" for. It does not, and is not
  meant to: `wide` is true for ANY entrypoint-validation error from ANY
  runtime, because the aggregate gate is all-or-nothing — an invalid Cline
  entrypoint reverts Codex's snapshot, which
  tests/configured-entrypoint-validation.test.cjs's 'an aggregate
  entrypoint validation failure rolls the Codex install back (#4249)'
  asserts directly. The discriminator is the error's KIND, not which
  runtime owns the failing path. Comment and CONTEXT.md now say that.

- Name the runtime in the "Configured entrypoint validation failed" error.
  ConfiguredEntrypointInvalid already carries `runtime`; the message threw
  it away, leaving an operator of a multi-runtime install unable to tell
  whose entrypoint broke — which matters precisely because the failure can
  revert a runtime that was itself fine.

- Set `configuredEntrypoints: []` explicitly on the copilot-instructions
  early return. Every other branch states the key; this one relied on
  installAllRuntimes' `(result.configuredEntrypoints || [])` defence.
  `[]` is correct, not a workaround: every Copilot hook is an inline
  printf one-liner (GSD_COPILOT_*_HOOK_BASH/PWSH), so there is no
  GSD-managed script or interpreter to resolve.

No behaviour change beyond the error-message text.

* docs(#4249): narrow the smoke scan's config-surface claim to what it checks

RUNTIME_CONFIG_FILES claimed every GSD-managed executable a runtime is
told to launch is registered in one of settings.json / hooks.json /
config.toml, and that nothing else in a config dir is runtime
configuration. Both halves are false as stated. Cline registers its hook
at .clinerules/hooks/PreToolUse — a subdirectory, and not one of those
names (writeClineArtifacts, src/runtime-hooks-surface.cts). Kimi's native
[[hooks]] config.toml lives under resolveKimiHooksTomlDir() (~/.kimi),
a directory separate from Kimi's own GSD configDir — the same gap
installer-migration 007 already documents as structurally unreachable.

The scan is in fact correct for what it runs against: entrypointRuntimes
defaults to claude + codex, whose launch paths do all live in those three
top-level files. Restate the docstring at that scope, name the two known
out-of-scope surfaces, and warn that adding either runtime to
entrypointRuntimes without teaching scanConfiguredEntrypoints about its
surface yields a scan that finds zero entrypoints and proves nothing.
The entrypointRuntimes default comment carried the same overgeneralization
("every other runtime reuses one of them") and is corrected with it.

Documentation only; no code change.

* fix(#4249): complete configuredEntrypoints/rollback shape on unparseable settings.local.json

An internal adversarial review (agy/gemini-3.8-flash-medium, round 8) found
that install()'s settings-json early return for an unparseable
settings.local.json omitted configuredEntrypoints and
rollbackInstallerMigrations from its result, unlike every other branch.
rollbackFinalizedInstallerMigrations reads result.rollbackInstallerMigrations
unconditionally, so this branch silently dropped its own installer-migration
rollback on a later finalize-stage failure.

Completed the return shape: configuredEntrypoints: [] (matching Copilot's
equally-early no-entrypoints-yet return) and rollbackInstallerMigrations
(already in closure scope). Red-then-green regression test added.

* test(#4249): ensure hooks/dist before packing in release-tarball-smoke.install.test.cjs

Same internal adversarial review (round 8): this suite's before() packed the
tarball directly, without the ensureHooksDist() guard every sibling
install-test suite (install.test.cjs, install-minimal-hooks.test.cjs,
mcp-catalog-parity.install.test.cjs) already uses. On a clean tree, or run
in isolation ahead of a suite that builds hooks/dist itself, this suite's
pack would ship a tarball with no hook scripts and fail closed on
SMOKE.INIT_FAILED instead of testing anything.

* fix(#4249): refresh stale test-timings weight for the codex-config split

next's own consolidation split (#4139/#4540) moved tests/codex-config.test.cjs's
heavy install()-pipeline blocks into tests/codex-config-hooks.test.cjs, but the
CI shard packer's weight table (tests/test-timings.json) was never updated:
codex-config.test.cjs still carried its pre-split weight (127783ms, ~18x the
suite mean), and codex-config-hooks.test.cjs — which now holds the #3245 block
this PR extends with its own #4249 install()-pipeline test — had no entry at
all, so the packer would silently underestimate it at the table's median
weight (roughly a 9x underestimate against its real cost).

trek-e's most recent review flagged a Windows shard timeout in-flight on
codex-config.test.cjs, plausibly aggravated by this PR's own addition to that
file before the rebase moved it. Re-measured both files locally (node --test
--test-reporter=tap, max of 3 runs, matching the table's own max-across-streams
methodology) and patched just these two entries — not a full regeneration,
which would need real multi-lane CI data this session doesn't have access to.

* fix(#4249): register configured-entrypoint-validation tests in the conformance-tier lists

next's platform-conformance-tier classifier (#4591/#4598) landed after this
branch's last rebase, so tests/configured-entrypoint-validation.test.cjs and
tests/codex-config-hooks.test.cjs were never classified, failing lint:ci's
gen-platform-conformance-tier --check and both the Linux and macOS conformance
suites.

* fix(#4249): drop codex-config.test.cjs from the #4733 pinned isolated-set expectation

next's #4733 (landed after this branch's last rebase) replaced the static
ISOLATED_HEAVY_FILES set with a threshold derived live from
tests/test-timings.json, and pins the current derived result in
EXPECTED_ISOLATED_UNIT_FILES for regression coverage. That pinned list still
named codex-config.test.cjs, whose own weight this PR already dropped from
127783ms to 189ms (after splitting its heavy install()-pipeline blocks into
codex-config-hooks.test.cjs) — well under #4733's derived 120000ms bar. The
live-computed set correctly no longer includes it; the pinned expectation is
updated to match.

* fix(#4249): name the rollback consequence in the entrypoint-validation error, and prove Cline's file survives it

trek-e's review flagged two Major gaps: the thrown error read identically
regardless of which of three real outcomes a runtime hit (nothing
persisted / snapshot reverted / config left broken on disk), and no test
proved the disclosed "left on disk, unreverted" case for Cursor/Windsurf/
Kimi/Cline — only Codex's revert path was ever asserted.

assertConfiguredEntrypoints now tags each invalid entry with its actual
consequence, mirrored from docs/how-to/update-gsd.md's existing
rollback-matrix disclosure. A new test drives the same aggregate failure
through Cline (whose own entrypoint is the one that fails) and asserts
its hook file is still on disk afterward.

* fix(#4249): close 4 gaps antigravity's adversarial review found in the entrypoint-validation PR

One review pass (gemini-3.8-flash-high via the antigravity review lane)
against this PR's full diff against next, findings independently verified
against source before fixing:

- Copilot's install() return object was the only one of 6 runtime branches
  missing rollbackInstallerMigrations — reachable now that this PR's own
  aggregate gate runs rollback across every result on any runtime's
  entrypoint failure, not just Copilot's own.
- buildHookCommand's unresolved-bash early return skipped track() entirely,
  so a win32 install with no Git Bash silently produced an unregistered
  .sh hook instead of the 'unresolved-interpreter' validation failure
  configuredEntrypointsForHook's own comment said it would.
- release-tarball-smoke.cjs reported a Cycle 4 install failure under
  SMOKE.INIT_FAILED (Cycle 1's code) instead of the already-existing
  SMOKE.INSTALL_FAILED.
- SCRIPT_PATH_RE excluded whitespace to avoid swallowing a shell command's
  trailing args, which also truncated any configDir containing a space
  (e.g. a real "/Users/John Doe/.claude"), silently zeroing the scan.
  Anchored the match on the already-known configDir prefix instead of a
  generic absolute-path guess: removes the ambiguity outright rather than
  patching the character class, and stays a raw-text scan on purpose (it
  catches a writer that emits a path without registering it — a
  JSON.parse of the expected schema would miss exactly that case).

One suggested finding (test-timings.json "missing" the new test file) was
verified false — that table only holds measured CI timings, populated
after a file's first real run — and one Ponytail suggestion (a
JSON.stringify dedup key) was rejected as it would reintroduce a real, if
narrow, key-collision risk for no benefit.

* fix(#4249): fix fork CI red from a stale changeset pr field and an unquoted docs/ comment

changeset-lint requires pr: to match the PR it runs on (16 on the fork,
not the eventual upstream number) — rehearsal-branch convention already
established earlier in this PR's history.

lint-docs-guard-registration's quote-pairing heuristic doesn't require the
docs/ path itself to be quoted — it flags a file once ANY quote-delimited
span containing "docs/" appears anywhere in it, alongside any real fs read
call. A comment ending "...update-gsd.md's rollback-matrix paragraph"
supplied the closing quote character (the possessive apostrophe) the
heuristic paired with an unrelated single-quoted string earlier in the
file. Reworded to avoid the unquoted apostrophe next to the path.

* chore(#4249): point the changeset pr field back at the upstream PR

Fork rehearsal (PR #16) is green; the real target for this changeset is
upstream PR #4249.

---------

Co-authored-by: Test <test@test.com>
Co-authored-by: Tom Boucher <trekkie@nomorestars.com>
This commit is contained in:
Dennis Alexis Valin Dittrich
2026-09-16 10:23:53 +02:00
committed by GitHub
parent 49f313d611
commit ad1477d659
17 changed files with 1515 additions and 72 deletions

View File

@@ -0,0 +1,5 @@
---
type: Changed
pr: 4249
---
**`gsd install` and `/gsd-update` now verify every GSD-managed runtime entrypoint before reporting success** — a hook script or its interpreter that is missing, unreadable, or not executable now fails the install with the offending paths named, instead of printing `Done!` over a configuration whose hooks can never fire.

View File

@@ -144,6 +144,14 @@ jobs:
if: steps.skip.outputs.skip != 'true'
run: npm ci
# `npm pack` runs prepack/prepare (build:lib) but NOT prepublishOnly, so
# a packed tarball is missing hooks/dist — the very launch targets the
# published package ships and the lifecycle smoke is meant to exercise.
# Build them here so the smoked tarball is publish-shaped (#4154).
- name: Build hooks (prepublishOnly parity)
if: steps.skip.outputs.skip != 'true'
run: npm run build:hooks
- name: Pack root tarball
if: steps.skip.outputs.skip != 'true'
id: pack

File diff suppressed because one or more lines are too long

View File

@@ -10850,6 +10850,7 @@ function install(isGlobal, runtime = DEFAULT_RUNTIME, options = {}) {
// Track installation failures
const failures = [];
const configuredEntrypoints = [];
let installerMigrationResult = null;
const rollbackInstallerMigrations = () => {
if (!installerMigrationResult || typeof installerMigrationResult.rollback !== 'function') return;
@@ -10932,12 +10933,17 @@ function install(isGlobal, runtime = DEFAULT_RUNTIME, options = {}) {
// destroyed by a wholesale delete whose snapshot lacked it.
let codexManagedSnapshotCaptured = false;
// null = the gate never ran; true/false = the gate ran and hooks/ (did|did
// not) exist as a directory pre-install. Three states are load-bearing: a
// not) exist as a directory pre-install. Two states are load-bearing: a
// clean first install records false, so its rollback removes the staged
// hooks/ tree entirely; minimal mode records null, so rollback does nothing.
// hooks/ tree entirely; a non-Codex runtime records null, so rollback does
// nothing.
let codexPreInstallHooksDirPreExisted = null;
let codexPreInstallHooksCaptureIncomplete = false;
if (_hostBehaviors(runtime).tomlConfigInstall && !isMinimalMode(_effectiveInstallMode)) {
// #4249 CR: not gated on install mode. restoreCodexSnapshot is reachable for
// a core/--minimal install too (#2695), and its pass-2 sweeps remove every
// gsd-* skill dir / agent file the snapshot does not claim — so an empty
// minimal-mode snapshot deleted the whole surface with nothing to restore.
if (_hostBehaviors(runtime).tomlConfigInstall) {
codexManagedSnapshotCaptured = true;
const _preSkillsDir = _resolveSkillsRootDir(runtime, targetDir, _installScopeId);
if (fs.existsSync(_preSkillsDir)) {
@@ -12824,6 +12830,7 @@ function install(isGlobal, runtime = DEFAULT_RUNTIME, options = {}) {
absoluteRunner: codexNodeRunner,
platform: process.platform,
});
configuredEntrypoints.push(...(hookWrite.configuredEntrypoints || []));
if (hookWrite.wrote) {
console.log(` ${green}✓${reset} Configured Codex hooks (SessionStart via hooks.json)`);
} else {
@@ -12903,7 +12910,21 @@ function install(isGlobal, runtime = DEFAULT_RUNTIME, options = {}) {
}
persistActiveProfileMarker();
return { settingsPath: null, settings: null, statuslineCommand: null, updateBannerCommand: null, runtime, configDir: targetDir };
// #4249: expose restoreCodexSnapshot (#3245) as a SECOND, separately named
// rollback rather than rebinding `rollbackInstallerMigrations` to it. A
// configured-entrypoint validation failure discovered later (outside this
// function, after Codex's own hooks.json/config.toml write already
// succeeded) previously had only the installer-migrations closure to call,
// leaving the just-written config.toml/hooks.json broken on disk despite
// Codex already owning a full pre-install snapshot/restore for exactly this.
//
// Every runtime's `rollbackInstallerMigrations` therefore still means what
// it says — the installer-migrations-only closure, which is what a
// finalize-stage failure that is NOT an entrypoint-validation failure gets
// (the Phase 4 contract). `rollbackPreInstallSnapshot` is Codex-only and is
// chosen only for entrypoint-validation failures. See the selection in
// installAllRuntimes' rollbackFinalizedInstallerMigrations.
return { settingsPath: null, settings: null, statuslineCommand: null, updateBannerCommand: null, runtime, configDir: targetDir, configuredEntrypoints, rollbackInstallerMigrations, rollbackPreInstallSnapshot: restoreCodexSnapshot };
}
if (plan.installSurface === 'copilot-instructions') {
@@ -12933,7 +12954,18 @@ function install(isGlobal, runtime = DEFAULT_RUNTIME, options = {}) {
writeCopilotHookConfig(targetDir);
console.log(` ${green}✓${reset} Configured Copilot lifecycle hook (sessionStart)`);
persistActiveProfileMarker();
return { settingsPath: null, settings: null, statuslineCommand: null, updateBannerCommand: null, runtime, configDir: targetDir };
// #4249: `[]`, not omitted — every Copilot hook is an inline `printf`
// one-liner (GSD_COPILOT_*_HOOK_BASH/PWSH in src/runtime-hooks-surface.cts),
// so this runtime genuinely launches no GSD-managed script and has no
// interpreter to resolve. Stated explicitly like every other branch rather
// than leaning on installAllRuntimes' `|| []` defence.
// #4249 (antigravity review): `rollbackInstallerMigrations` was missing here
// — every other branch returns it. This PR's own aggregate entrypoint gate
// is what makes the gap reachable: an unrelated runtime's invalid entrypoint
// now triggers rollbackFinalizedInstallerMigrations for every result in the
// batch, and a Copilot result with no rollback function silently skips
// reverting Copilot's own installer migrations.
return { settingsPath: null, settings: null, statuslineCommand: null, updateBannerCommand: null, runtime, configDir: targetDir, configuredEntrypoints: [], rollbackInstallerMigrations };
}
if (plan.installSurface === 'cursor-hooks-json') {
@@ -12956,7 +12988,7 @@ function install(isGlobal, runtime = DEFAULT_RUNTIME, options = {}) {
// The re-run is retained for parity with the settings.json install path.
writeManifest(targetDir, runtime, { mode: _effectiveInstallMode, scope: _installScopeId });
persistActiveProfileMarker();
return { settingsPath: null, settings: null, statuslineCommand: null, updateBannerCommand: null, runtime, configDir: targetDir };
return { settingsPath: null, settings: null, statuslineCommand: null, updateBannerCommand: null, runtime, configDir: targetDir, configuredEntrypoints: cursorHookResult.configuredEntrypoints, rollbackInstallerMigrations };
}
if (plan.installSurface === 'profile-marker-only') {
@@ -13012,6 +13044,7 @@ function install(isGlobal, runtime = DEFAULT_RUNTIME, options = {}) {
const kimiHookOpts = { portableHooks: hasPortableHooks, runtime };
const kimiHooksTomlPath = path.join(kimiHooksRoot, 'config.toml');
const kimiHooksResult = writeKimiHooksToml(kimiHooksTomlPath, kimiHooksRoot, { hookOpts: kimiHookOpts });
configuredEntrypoints.push(...kimiHooksResult.configuredEntrypoints);
if (kimiHooksResult.changed) {
console.log(` ${green}✓${reset} Configured ${kimiHooksResult.entryCount} GSD hook(s) in ${kimiHooksTomlPath}`);
}
@@ -13068,6 +13101,7 @@ function install(isGlobal, runtime = DEFAULT_RUNTIME, options = {}) {
const windsurfHookResult = writeWindsurfHooksJson(targetDir, src, {
platform: process.platform,
});
configuredEntrypoints.push(...windsurfHookResult.configuredEntrypoints);
if (windsurfHookResult.changed) {
console.log(` ${green}✓${reset} Configured Windsurf lifecycle hooks (pre_write_code, pre_run_command)`);
} else {
@@ -13083,19 +13117,19 @@ function install(isGlobal, runtime = DEFAULT_RUNTIME, options = {}) {
}
persistActiveProfileMarker();
return { settingsPath: null, settings: null, statuslineCommand: null, updateBannerCommand: null, runtime, configDir: targetDir };
return { settingsPath: null, settings: null, statuslineCommand: null, updateBannerCommand: null, runtime, configDir: targetDir, configuredEntrypoints, rollbackInstallerMigrations };
}
if (plan.installSurface === 'cline-rules') {
// Cline uses the `.clinerules/` directory form (issue #787): GSD rules live
// at .clinerules/gsd.md and a PreToolUse lifecycle hook at
// .clinerules/hooks/PreToolUse. Global installs also get ~/.agents/AGENTS.md.
writeClineArtifacts(targetDir, isGlobal);
const clineArtifacts = writeClineArtifacts(targetDir, isGlobal);
// Re-run the manifest pass: these artifacts are written *after* the earlier
// writeManifest() call, so a second pass is needed to hash-track them.
writeManifest(targetDir, runtime, { mode: _effectiveInstallMode, scope: _installScopeId });
persistActiveProfileMarker();
return { settingsPath: null, settings: null, statuslineCommand: null, updateBannerCommand: null, runtime, configDir: targetDir };
return { settingsPath: null, settings: null, statuslineCommand: null, updateBannerCommand: null, runtime, configDir: targetDir, configuredEntrypoints: clineArtifacts.configuredEntrypoints, rollbackInstallerMigrations };
}
// Configure statusline and hooks in settings.json (or settings.local.json for local Claude installs).
@@ -13220,8 +13254,11 @@ function install(isGlobal, runtime = DEFAULT_RUNTIME, options = {}) {
persistActiveProfileMarker();
// Callers index this result by `runtime` (installAllRuntimes' statusline
// lookup), so every early exit must return the full shape — a bare return
// crashes the install rather than skipping one file.
return { settingsPath: null, settings: null, statuslineCommand: null, updateBannerCommand: null, runtime, configDir: targetDir };
// crashes the install rather than skipping one file. That includes
// configuredEntrypoints/rollbackInstallerMigrations: rollbackFinalizedInstallerMigrations
// reads result.rollbackInstallerMigrations unconditionally, and an omitted
// field there silently skips this runtime's rollback on a finalize-stage failure.
return { settingsPath: null, settings: null, statuslineCommand: null, updateBannerCommand: null, runtime, configDir: targetDir, configuredEntrypoints: [], rollbackInstallerMigrations };
}
const settings = validateHookFields(cleanupOrphanedHooks(rawSettings));
// #3002 CR / #3662: rewrite legacy `node .../gsd-*.js` command strings (pre-
@@ -13243,7 +13280,19 @@ function install(isGlobal, runtime = DEFAULT_RUNTIME, options = {}) {
// runtime's hostBehaviors instead of a hardcoded `runtime === 'antigravity'`
// check inside projectLocalHookPrefix.
const localPrefix = projectLocalHookPrefix({ runtime, dirName, hookPathStyle: _hostBehaviors(runtime).hookPathStyle });
const hookOpts = { portableHooks: hasPortableHooks, runtime };
const settingsEntrypoints = [];
const hookOpts = {
portableHooks: hasPortableHooks,
runtime,
configPath: settingsPath,
// #4249: track unconditionally. Gating on `plan.hooksSurface ===
// 'settings-json'` made tracking depend on an unasserted
// installSurface/hooksSurface coupling — a descriptor that broke it would
// silently drop this runtime out of validation. Everything recorded here
// lands in settings.json by construction, and the registered-command
// filter below already discards entries no hook actually references.
configuredEntrypoints: settingsEntrypoints,
};
// #2979: local-install hook commands also use a runner GUI/minimal-PATH
// runtimes can resolve. Bare `node` fails when the host launches the
// runtime with a stripped PATH (Finder/Antigravity/etc) — #3662 replaces
@@ -13257,19 +13306,19 @@ function install(isGlobal, runtime = DEFAULT_RUNTIME, options = {}) {
// `node` command that recreates the #2979 failure.
const localCmd = (hookFile) => localNodeRunner === null
? null
: projectShellCommandText({
: hooksSurface.recordConfiguredHookCommand(projectShellCommandText({
runnerToken: localNodeRunner,
argTokens: [`${localPrefix}/hooks/${hookFile}`],
runtime,
platform: process.platform,
});
const localShellCmd = (hookFile) => buildLocalShellHookCommand({
}), targetDir, hookFile, hookOpts);
const localShellCmd = (hookFile) => hooksSurface.recordConfiguredHookCommand(buildLocalShellHookCommand({
localPrefix,
hookFile,
bashRunner: localBashRunner,
runtime,
platform: process.platform,
});
}), targetDir, hookFile, hookOpts);
const statuslineCommand = isGlobal
? buildHookCommand(targetDir, 'gsd-statusline.js', hookOpts)
: localCmd('gsd-statusline.js');
@@ -13339,6 +13388,31 @@ function install(isGlobal, runtime = DEFAULT_RUNTIME, options = {}) {
? buildHookCommand(targetDir, 'gsd-update-banner.js', hookOpts)
: localCmd('gsd-update-banner.js'));
const registeredHookCommands = Object.values(settings.hooks || {})
.flatMap(groups => Array.isArray(groups) ? groups : [])
.flatMap(group => Array.isArray(group && group.hooks) ? group.hooks : [])
.map(hook => hook && hook.command)
.filter(command => typeof command === 'string');
// #4249: match by the managed script's `/hooks/<basename>` path segment, not
// by exact command-string equality. The blocking-guard hooks above register
// only-if-absent, so a hook already present from a prior install keeps its
// OLD command untouched — but `track()` always records the FRESHLY computed
// command for it, which never equals what's actually persisted. Matching on
// the segment (present in the persisted command either way, since every
// entry.scriptPath is <configDir>/hooks/<name> by construction) keeps an
// already-registered, still-active hook in the validated set instead of
// silently dropping it (#4154 Blocker) — anchored on `/hooks/` rather than a
// bare basename so an unrelated user command that merely mentions the same
// filename can't false-positive into GSD's validated set.
configuredEntrypoints.push(
...settingsEntrypoints.filter(entry => {
const hooksSegment = '/hooks/' + path.basename(entry.scriptPath);
return registeredHookCommands.some(command => command.includes(hooksSegment));
}),
);
const statuslineEntrypoints = settingsEntrypoints.filter(entry => entry.command === statuslineCommand);
const updateBannerEntrypoints = settingsEntrypoints.filter(entry => entry.command === updateBannerCommand);
// #683: Set worktree.baseRef:"head" in settings.local.json for local Claude installs.
// Both fresh and upgrade paths apply only when worktrees are enabled for the project.
// Never applies to global installs, non-Claude runtimes, or when the user already
@@ -13407,15 +13481,65 @@ function install(isGlobal, runtime = DEFAULT_RUNTIME, options = {}) {
settings,
statuslineCommand,
updateBannerCommand,
statuslineEntrypoints,
updateBannerEntrypoints,
runtime,
configDir: targetDir,
rollbackInstallerMigrations,
configuredEntrypoints,
};
}
/**
* Apply statusline config, then print completion message
*/
// #4249 (review, Major): rollback consequence differs by runtime surface —
// see docs/how-to/update-gsd.md's rollback-matrix paragraph, which this
// mirrors. Codex reverts (pre-install snapshot restore); Cursor/Windsurf/
// Kimi/Kimi Code/Cline already wrote their config file inside install(),
// ahead of this gate, with no revert path, so it is left on disk broken;
// every other (settings.json-based) runtime writes strictly after this gate,
// so a failure here means nothing new was persisted for it.
const ENTRYPOINT_LEFT_UNREVERTED_RUNTIMES = new Set(['cursor', 'windsurf', 'kimi', 'kimi-code', 'cline']);
function describeEntrypointConsequence(invalidRuntime) {
if (invalidRuntime === 'codex') return 'reverted: its pre-install snapshot was restored';
if (ENTRYPOINT_LEFT_UNREVERTED_RUNTIMES.has(invalidRuntime)) return 'NOT reverted: its config file is already written and was left on disk — fix the reported path and rerun install';
return 'not persisted: this runtime writes its config after this check';
}
function assertConfiguredEntrypoints(entries) {
// #4249: some writers push the same (configPath, scriptPath) pair more than
// once (e.g. Kimi's context-monitor hook registered under several events,
// or the portable resolver script shared by every portable JS hook) — keep
// one so a broken entry is reported once, not once per duplicate.
const seen = new Set();
const deduped = (entries || []).filter((entry) => {
const key = JSON.stringify([entry.configPath, entry.scriptPath]);
if (seen.has(key)) return false;
seen.add(key);
return true;
});
const validation = hooksSurface.validateConfiguredEntrypoints(deduped);
if (validation.ok) return;
const error = new Error(
// #4249: lead each entry with its runtime, and name the actual consequence
// for that runtime (review, Major) — the aggregate gate is all-or-nothing
// across every runtime being installed, and a failure here can revert a
// runtime whose own entrypoints were fine (see
// rollbackFinalizedInstallerMigrations) while leaving another runtime's
// already-written config broken on disk with no revert at all, so an
// operator reading only this message must be able to tell WHOSE
// entrypoint broke and WHAT that means for their config, not just that
// something did.
`Configured entrypoint validation failed: ${validation.invalid.map(({ runtime: invalidRuntime, role, path: invalidPath, reason }) => `${invalidRuntime} ${role} ${invalidPath} (${reason}) [${describeEntrypointConsequence(invalidRuntime)}]`).join(', ')}`,
);
error.configuredEntrypointValidation = validation;
throw error;
}
// #4249: `bannerOpts.configuredEntrypoints` is the ONLY source assertConfiguredEntrypoints
// checks below — a caller that omits it (or calls finishInstall directly instead of
// through installAllRuntimes) gets zero entrypoint validation, silently. installAllRuntimes
// always passes the full set (per-runtime entries plus statusline/updateBanner); any other
// caller must do the same for this gate to mean anything.
function finishInstall(settingsPath, settings, statuslineCommand, shouldInstallStatusline, runtime = DEFAULT_RUNTIME, isGlobal = true, configDir = null, bannerOpts = {}) {
// #2093: isKilo dropped — the Kilo permissions-writer call below is gated
// on plan.finishPermissionWriter === 'kilo' (descriptor-driven), not this flag.
@@ -13429,6 +13553,19 @@ function finishInstall(settingsPath, settings, statuslineCommand, shouldInstallS
const { isOpencode, isCodex, isCursor, isAugment, isQwen, isHermes, isCline } = runtimeFlags(runtime);
const plan = resolveInstallPlan(runtime);
// #4249 Major: validate BEFORE this function's own settings.json write (and
// before writeNonClaudeDefaults) instead of after. Cursor/Windsurf/Kimi/Cline
// already persisted their config inside install() by this point, with no
// rollback path covering those writes; Codex also persists inside install()
// but its rollback binds to a full pre-install snapshot restore, so it IS
// covered (see docs/how-to/update-gsd.md). For the settings-json surface
// this ordering means a failing validation never reaches this function's
// own write at all. On the production path this is a redundant backstop —
// installAllRuntimes's own aggregate assertConfiguredEntrypoints call
// already validates the superset before finishInstall runs for any
// runtime — kept for a caller that invokes finishInstall directly.
assertConfiguredEntrypoints(bannerOpts.configuredEntrypoints);
if (shouldInstallStatusline && plan.writesSharedSettings && !_hostBehaviors(runtime).skipSettingsUi) {
if (!isGlobal && !forceStatusline) {
// Local installs skip statusLine by default: repo settings.json takes precedence over
@@ -14326,10 +14463,32 @@ function installAllRuntimes(runtimes, isGlobal, isInteractive) {
const rollbackFinalizedInstallerMigrations = (error) => {
const rollbackFailures = [];
// #4249: this discriminates on the error's KIND, never on which runtime
// owns the failing entrypoint. `wide` is true for ANY entrypoint-validation
// failure from ANY runtime, by design: the aggregate gate exists so a
// multi-runtime install cannot report success while one of its entrypoints
// is broken, so an invalid Cline entrypoint reverts Codex's pre-install
// snapshot too — even though Codex itself was fine and its own "Done!"
// summary already printed. tests/configured-entrypoint-validation.test.cjs
// ('an aggregate entrypoint validation failure rolls the Codex install
// back') exercises exactly that, and it is the all-or-nothing behaviour
// docs/how-to/update-gsd.md documents.
//
// What this narrows is the OTHER axis: a finalize-stage exception that is
// not an entrypoint-validation failure at all — e.g. a sibling runtime's
// permission-config write dying with EACCES — gets only the
// installer-migrations-only rollback that Phase 4 specifies
// (docs/installer-migrations.md#phase-4-installupdate-integration).
// Un-installing (and, on update, downgrading) an already-"Done!" Codex over
// an unrelated error is not an outcome any doc promises, while the sibling
// surfaces that write config inside install() would keep theirs regardless.
const wide = !!(error && error.configuredEntrypointValidation);
for (const result of [...results].reverse()) {
if (!result || typeof result.rollbackInstallerMigrations !== 'function') continue;
if (!result) continue;
const rollback = (wide && result.rollbackPreInstallSnapshot) || result.rollbackInstallerMigrations;
if (typeof rollback !== 'function') continue;
try {
result.rollbackInstallerMigrations();
rollback();
} catch (rollbackError) {
rollbackFailures.push({
runtime: result.runtime,
@@ -14357,6 +14516,19 @@ function installAllRuntimes(runtimes, isGlobal, isInteractive) {
const finalize = (shouldInstallStatusline, shouldInstallBanner) => {
try {
const selectedConfiguredEntrypoints = (result) => {
if (!result || result.skipped) return [];
const useStatusline = statuslineRuntimes.includes(result.runtime)
&& shouldInstallStatusline
&& (isGlobal || forceStatusline);
return [
...(result.configuredEntrypoints || []),
...(useStatusline ? (result.statuslineEntrypoints || []) : []),
...(shouldInstallBanner ? (result.updateBannerEntrypoints || []) : []),
];
};
assertConfiguredEntrypoints(results.flatMap(selectedConfiguredEntrypoints));
const printSummaries = () => {
for (const result of results) {
if (result && result.skipped) continue;
@@ -14370,7 +14542,11 @@ function installAllRuntimes(runtimes, isGlobal, isInteractive) {
result.runtime,
isGlobal,
result.configDir,
{ shouldInstallBanner: !!shouldInstallBanner, bannerCommand: result.updateBannerCommand }
{
shouldInstallBanner: !!shouldInstallBanner,
bannerCommand: result.updateBannerCommand,
configuredEntrypoints: selectedConfiguredEntrypoints(result),
}
);
}
};

View File

@@ -476,7 +476,7 @@ Node.js CLI utility (`gsd-tools.cjs`) with domain modules split across `gsd-core
| `graphify-command-router.cjs` | ADR-959 capability command router — first real capability command cutover (phase 4d-impl-2); extracted from the `case 'graphify':` arm in `gsd-tools.cjs`; dispatches build/query/status/diff subcommands; discovered via `commandFamilies` in the capability registry |
| `audit-command-router.cjs` | ADR-959 capability command router (phase 4d-impl-3); extracted from the `case 'audit-uat':` and `case 'audit-open':` arms in `gsd-tools.cjs`; `routeAuditUat` → `uat.cjs:cmdAuditUat`, `routeAuditOpen` → `audit.cjs:{auditOpenArtifacts,formatAuditReport}`; discovered via `commandFamilies` in the capability registry |
| `intel-command-router.cjs` | ADR-959 capability command router (phase 4d-impl-4, last first-party cutover); extracted from the `case 'intel':` arm in `gsd-tools.cjs`; `routeIntelCommand` → all 9 intel subcommands via lazy `require('./intel.cjs')`; preserves non-raw `timeAgo` transform on `status.files[*].updated_at`; discovered via `commandFamilies` in the capability registry |
| `runtime-hooks-surface.cjs` | Standalone hook-surface writer module (ADR-857 phase 5f-1); owns Cline rules/agents-md/pre-tool-use hook generation, Cursor `hooks.json` reconciliation, Copilot session-hook config, and Codex hook-block management; extracted verbatim from `bin/install.js` with no logic change. |
| `runtime-hooks-surface.cjs` | Hook-surface writer and configured-entrypoint validation module (ADR-857 phase 5f-1); owns Cline rules/agents-md/pre-tool-use hook generation, Cursor/Windsurf `hooks.json`, Kimi hook TOML, Copilot session-hook config, Codex hook-block management, and construction-time records for paths emitted into runtime configuration. Installer success requires those records to pass non-executing file-type and interpreter-resolution checks. |
---

View File

@@ -26,7 +26,11 @@ GSD will:
8. Offer to restore the user-added files it backed up in step 5.
9. Report whether locally modified GSD files were backed up to `gsd-local-patches/`.
Restart your runtime after the update to pick up new commands and agents.
Before reporting completion, the installer checks each GSD-managed script and interpreter path written into runtime configuration, resolving the interpreter against the current install-time `PATH`. This only proves the path resolves now — a hook fired later under a different, more restricted `PATH` (e.g. a GUI launcher) can still fail even after this check passes.
If a script is missing, unreadable, has the wrong file type, lacks a required execute permission, or its interpreter cannot be resolved, the update fails and reports every invalid path, each tagged with what happens to that runtime's config next. For Claude Code and other settings.json-based runtimes, this check runs before the update writes settings.json, so nothing new is persisted (an earlier settings.json/settings.local.json migration, if one applied, is the one exception and stays applied) — reported as "not persisted". For Codex, this reverts config.toml/hooks.json along with the rest of that runtime's pre-install snapshot (skills/, agents/, gsd-core/VERSION) — reported as "reverted". For Cursor, Windsurf, Kimi, and Cline, the runtime's config file is already written earlier in the update, ahead of this check, so a failure is reported but that file is left in place, broken — reported as "NOT reverted". Fix the reported path problem and rerun `/gsd-update` — do not restart into the incomplete update.
Restart your runtime after a successful update to pick up new commands and agents.
---

View File

@@ -52,6 +52,7 @@ module.exports = {
"tests/config-loader.test.cjs",
"tests/config-schema.property.test.cjs",
"tests/config.test.cjs",
"tests/configured-entrypoint-validation.test.cjs",
"tests/contributor-standards.test.cjs",
"tests/core-utils.test.cjs",
"tests/cursor-subagent-isolation.test.cjs",

View File

@@ -62,6 +62,7 @@ module.exports = {
"tests/config-loader.test.cjs",
"tests/config-schema.property.test.cjs",
"tests/config.test.cjs",
"tests/configured-entrypoint-validation.test.cjs",
"tests/copilot-install.test.cjs",
"tests/copilot-upgrades.test.cjs",
"tests/core-utils.test.cjs",

View File

@@ -37,6 +37,12 @@
* colon-namespace leaks (WORKFLOW_BODY_COLON_LEAK).
* This check populates result.details with counters but does NOT return a
* failure code by default; it is informational until enforcement is enabled.
*
* Configured-entrypoint checks (Cycle 4 — #4154):
* For each runtime in entrypointRuntimes, runs the tarball-installed
* installer into a throwaway HOME, then re-reads that runtime's own written
* config files and asserts every GSD-managed script path they name resolves
* to a file (ENTRYPOINT_UNRESOLVED). Requires fixtureDir.
*/
'use strict';
@@ -47,6 +53,8 @@ const os = require('os');
const path = require('path');
const { PACKAGE_NAME } = require('../gsd-core/bin/lib/package-identity.cjs');
const { ExitError, runMain } = require('./lib/cli-exit.cjs');
const shellCmdProjection = require('../gsd-core/bin/lib/shell-command-projection.cjs');
const { escapeRegex } = require('../gsd-core/bin/lib/pattern.cjs');
// 120 s proved too tight for cold-cache `npm install -g` of a 1499-file tarball:
// spawnSync fires SIGTERM at the deadline and returns { status: null, stdout: '',
// stderr: '' } (Node docs: status is null when a subprocess is terminated by a
@@ -86,6 +94,8 @@ const SMOKE = Object.freeze({
INIT_FAILED: 'init_failed',
// Cycle 3 code
WORKFLOW_BODY_COLON_LEAK: 'workflow_body_colon_leak',
// Cycle 4 code (#4154)
ENTRYPOINT_UNRESOLVED: 'entrypoint_unresolved',
});
// ---------------------------------------------------------------------------
@@ -274,6 +284,106 @@ function scanWorkflowColonLeak(filePath, cmdNames) {
return null;
}
// ---------------------------------------------------------------------------
// Cycle 4 helpers: configured-entrypoint resolution (#4154)
// ---------------------------------------------------------------------------
/**
* Top-level config files this scan reads. These are the surfaces that carry
* every GSD-managed launch path for the runtimes Cycle 4 actually installs
* (`entrypointRuntimes`, default claude + codex): claude registers into
* settings.json, codex into hooks.json and config.toml.
*
* This is NOT an exhaustive map of where GSD writes launch paths across all
* runtimes, and the scan is top-level only by design. Two known surfaces sit
* outside it: Cline registers its hook at `.clinerules/hooks/PreToolUse` (a
* subdirectory, and not one of these names — see writeClineArtifacts in
* src/runtime-hooks-surface.cts), and Kimi's native `[[hooks]]` config.toml
* lives under `resolveKimiHooksTomlDir()` (`~/.kimi`), a directory separate
* from Kimi's own GSD configDir. Adding either runtime to entrypointRuntimes
* requires teaching scanConfiguredEntrypoints about its surface first,
* otherwise the scan reports zero entrypoints and silently proves nothing.
*/
const RUNTIME_CONFIG_FILES = Object.freeze(['settings.json', 'hooks.json', 'config.toml']);
/**
* Extract every script path `text` names underneath `configDir`.
*
* #4249 (antigravity review): anchored on the literal, already-known
* `configDir` prefix instead of a generic "any absolute path" character
* class. The prior version excluded whitespace from the match to avoid
* swallowing a shell command's trailing args, which also truncated any
* legitimate path containing a space (e.g. `/Users/John Doe/.claude`) —
* `scanConfiguredEntrypoints` would then silently report zero checked
* paths. Anchoring on `configDir` removes the ambiguity outright: a match
* can only start where the known prefix literally occurs in the text, so
* an interpreter path concatenated ahead of it (`"/usr/bin/node
* /configDir/hooks/foo.js"`) is never swallowed either, and interior
* whitespace inside `configDir` or the script's own path segments is safe
* to allow. This still scans raw, unparsed config text on purpose (see
* scanConfiguredEntrypoints's doc comment) — it catches a writer that
* embeds a launch path without registering it, which a structured
* JSON.parse of the expected schema would miss entirely.
*
* Windows configs store paths with backslashes, which JSON/TOML doubles on
* write; collapsing `\\` to `\` first makes the raw text scan work on both
* platforms without parsing each config format separately (POSIX text has no
* backslashes, so the collapse is a no-op there).
*
* Every writer bakes `configDir` through the same posixNormalize seam
* (src/runtime-hooks-surface.cts) before writing it into config text, on
* every platform — so the anchor must match that projection, not the
* OS-native `configDir` string this function receives.
*/
function configuredEntrypointsIn(text, configDir) {
const normalizedPrefix = shellCmdProjection.posixNormalize(configDir).replace(/\/+$/, '') + '/';
const scriptPathRe = new RegExp(`${escapeRegex(normalizedPrefix)}[^"']{0,400}?\\.(?:js|cjs|mjs|sh|cmd|ps1)`, 'g');
const found = new Set();
for (const match of text.replace(/\\\\/g, '\\').matchAll(scriptPathRe)) {
found.add(path.resolve(match[0]));
}
return [...found];
}
/**
* Throwaway HOME the Cycle 4 install for `runtime` runs against. Exported so a
* test can seed that runtime's config before the install rather than
* hard-coding the layout runSmoke picks.
*/
function entrypointFixtureHome(fixtureDir, runtime) {
return path.join(fixtureDir, `entrypoints-${runtime}`);
}
/**
* Re-derive the configured entrypoints a completed install wrote into a
* runtime's own config files and report the ones that do not resolve to a
* file. Internal to Cycle 4; the smoke's verdict is the exported contract.
*
* This deliberately does NOT consult the installer's own entrypoint list (the
* assertConfiguredEntrypoints gate in bin/install.js). That gate can only
* validate paths a config writer remembered to register; reading the written
* config back is what catches a writer that emits a launch path without
* registering it, and a stale registration an install left behind.
*
* @param {string} configDir - Absolute path to the runtime config dir.
* @returns {{ checked: string[], unresolved: { configPath: string, scriptPath: string }[] }}
*/
function scanConfiguredEntrypoints(configDir) {
const checked = [];
const unresolved = [];
for (const name of RUNTIME_CONFIG_FILES) {
const configPath = path.join(configDir, name);
if (!fs.existsSync(configPath) || !fs.statSync(configPath).isFile()) continue;
const text = fs.readFileSync(configPath, 'utf-8');
for (const scriptPath of configuredEntrypointsIn(text, configDir)) {
checked.push(scriptPath);
if (fs.existsSync(scriptPath) && fs.statSync(scriptPath).isFile()) continue;
unresolved.push({ configPath, scriptPath });
}
}
return { checked, unresolved };
}
// ---------------------------------------------------------------------------
// Pure function: runSmoke
// ---------------------------------------------------------------------------
@@ -285,6 +395,8 @@ function scanWorkflowColonLeak(filePath, cmdNames) {
* @param {string} opts.expectedVersion - semver string to assert (e.g. "1.50.0")
* @param {string} [opts.fixtureDir] - Temp dir to run `init` into (must NOT be HOME)
* @param {string[]} [opts.lifecycleCommands] - Commands to file-check (default: see below)
* @param {string[]} [opts.entrypointRuntimes] - Runtime profiles whose configured entrypoints are
* re-checked after a real install (default: see below). Requires fixtureDir; pass [] to skip.
* @param {boolean} [opts.dryRun=false] - If true, skip actual npm install; validate input only
* @param {object} [opts.npmEnv] - Optional env dict for the internal npm install
* spawnSync call. Pass an isolated HOME env (e.g. from isolatedNpmEnv() in tests/helpers.cjs)
@@ -298,6 +410,11 @@ function runSmoke({
expectedVersion,
fixtureDir,
lifecycleCommands = ['init', 'discuss-phase', 'plan-phase', 'execute-phase'],
// claude and codex cover the two top-level config surfaces this scan knows
// how to read (settings.json, and hooks.json + config.toml). Most other
// runtimes reuse one of those two shapes; Cline and Kimi do not (see
// RUNTIME_CONFIG_FILES), so they are out of scope here rather than covered.
entrypointRuntimes = ['claude', 'codex'],
dryRun = false,
npmEnv = undefined,
}) {
@@ -551,6 +668,75 @@ function runSmoke({
// NOTE: colonLeakCount is informational here. Once the backlog is fixed,
// a future enforcement mode can fail on non-zero counts.
// ─────────────────────────────────────────────────────────────────────────
// Cycle 4: configured-entrypoint resolution (#4154)
// ─────────────────────────────────────────────────────────────────────────
// The installer's own gate (assertConfiguredEntrypoints) runs in-process and
// only sees the paths a config writer registered with it. Installing the
// packed tarball for real and reading each runtime's written config back is
// what proves the launch paths a user's runtime will actually invoke exist
// in the shipped layout.
const entrypointProfiles = [];
if (fixtureDir && entrypointRuntimes.length > 0) {
for (const runtime of entrypointRuntimes) {
// Own HOME per runtime so a --global install cannot reach the real one.
const runtimeHome = entrypointFixtureHome(fixtureDir, runtime);
const configDir = path.join(runtimeHome, `.${runtime}`);
fs.mkdirSync(runtimeHome, { recursive: true });
const installResult = spawnSync(
process.execPath,
[path.join(pkg, 'bin', 'install.js'), `--${runtime}`, '--global', '--config-dir', configDir],
{
encoding: 'utf-8',
cwd: runtimeHome,
stdio: ['pipe', 'pipe', 'pipe'],
env: {
...effectiveNpmEnv,
HOME: runtimeHome,
USERPROFILE: runtimeHome,
// Same reason as the init check above: install.js skips its main()
// block when GSD_TEST_MODE is set and would exit 0 writing nothing.
GSD_TEST_MODE: '',
NO_UPDATE_NOTIFIER: '1',
},
timeout: CHILD_TIMEOUT_MS,
},
);
if (installResult.status !== 0) {
// #4249 (antigravity review): this is the Cycle 4 per-runtime install,
// not Cycle 1's `gsd init` — SMOKE.INSTALL_FAILED is the code Cycle 2's
// identical spawnSync-failure check already uses for the same failure
// class; reusing SMOKE.INIT_FAILED here conflated the two lifecycle
// stages in the reported code.
return {
code: SMOKE.INSTALL_FAILED,
details: {
...details,
runtime,
configDir,
stderr: installResult.stderr,
stdout: installResult.stdout,
},
};
}
const scan = scanConfiguredEntrypoints(configDir);
if (scan.unresolved.length > 0) {
return {
code: SMOKE.ENTRYPOINT_UNRESOLVED,
details: { ...details, runtime, configDir, unresolved: scan.unresolved },
};
}
entrypointProfiles.push({ runtime, configDir, entrypointsChecked: scan.checked.length });
}
}
details.entrypointProfiles = entrypointProfiles;
return { code: SMOKE.OK, details };
}
@@ -635,7 +821,14 @@ function cleanup(...dirs) {
// Exports
// ---------------------------------------------------------------------------
module.exports = { SMOKE, runSmoke, binInvocation, CHILD_TIMEOUT_MS };
module.exports = {
SMOKE,
runSmoke,
binInvocation,
entrypointFixtureHome,
CHILD_TIMEOUT_MS,
configuredEntrypointsIn,
};
if (require.main === module) {
runMain(cliMain);

View File

@@ -56,6 +56,7 @@ const {
shellHookOmitsBashRunner,
escapeTomlDoubleQuotedString,
escapePosixDoubleQuoted,
resolveExecutableBinary,
} = shellCmdProjection as {
isManagedHookBasename: (scriptPath: string, opts?: { surface?: string }) => boolean;
isManagedHookCommand: (cmd: string | null | undefined, opts?: { surface?: string; includeLegacyAliases?: boolean; configDir?: string }) => boolean;
@@ -66,6 +67,7 @@ const {
shellHookOmitsBashRunner: (opts: { platform: string; runtime: string; isShellHook: boolean }) => boolean;
escapeTomlDoubleQuotedString: (value: unknown) => string;
escapePosixDoubleQuoted: (value: unknown) => string;
resolveExecutableBinary: (name: string, opts?: { platform?: string; requireExecutable?: boolean }) => string | null;
};
// ---------------------------------------------------------------------------
@@ -642,7 +644,7 @@ interface BashRunnerOpts {
existsSync?: (p: string) => boolean;
}
function resolveBashRunner(opts?: BashRunnerOpts): string | null {
function resolveBashExecutable(opts?: BashRunnerOpts): string | null {
const platform = (opts && opts.platform) || process.platform;
if (platform !== 'win32') return 'bash';
@@ -658,13 +660,19 @@ function resolveBashRunner(opts?: BashRunnerOpts): string | null {
}
for (const candidate of candidates) {
if (candidate && exists(candidate)) {
return JSON.stringify(shellCmdProjection.posixNormalize(candidate));
}
if (candidate && exists(candidate)) return shellCmdProjection.posixNormalize(candidate);
}
return null;
}
function resolveBashRunner(opts?: BashRunnerOpts): string | null {
const executable = resolveBashExecutable(opts);
if (executable === null) return null;
return ((opts && opts.platform) || process.platform) === 'win32'
? JSON.stringify(executable)
: executable;
}
// ---------------------------------------------------------------------------
// Shared: rewriteLegacyManagedNodeHookCommands
// ---------------------------------------------------------------------------
@@ -929,6 +937,7 @@ interface ReconcileResult {
changed: boolean;
wrote: boolean;
path: string;
configuredEntrypoints?: ConfiguredEntrypoint[];
}
/**
@@ -1117,14 +1126,17 @@ interface ShimIR {
render: { cmd: () => string };
}
function parseAbsoluteRunnerToken(absoluteRunnerToken: string): string {
try {
return JSON.parse(absoluteRunnerToken) as string;
} catch {
return absoluteRunnerToken;
}
}
function buildCodexHookWindowsShimIR(scriptAbsPath: string, absoluteRunnerToken: string | null): ShimIR | null {
if (!absoluteRunnerToken) return null;
let interpreter: string;
try {
interpreter = JSON.parse(absoluteRunnerToken) as string;
} catch {
interpreter = absoluteRunnerToken;
}
const interpreter = parseAbsoluteRunnerToken(absoluteRunnerToken);
const targetAbs = shellCmdProjection.posixNormalize(scriptAbsPath);
const scriptQuoted = JSON.stringify(targetAbs);
const cmdPath = scriptAbsPath.replace(/\.js$/, '.cmd');
@@ -1159,8 +1171,9 @@ function ensureCodexHooksJsonSessionStart(targetDir: string, opts: EnsureCodexSe
const scriptPath = shellCmdProjection.posixNormalize(path.resolve(targetDir, 'hooks', 'gsd-check-update.js'));
const cmdShimPath = scriptPath.replace(/\.js$/, '.cmd');
const configuredEntrypoints: ConfiguredEntrypoint[] = [];
let managedCommand: string | undefined;
if (platform === 'win32') {
const shimIR = buildCodexHookWindowsShimIR(scriptPath, absoluteRunner);
if (!shimIR) return { changed: false, wrote: false, path: hooksJsonPath };
@@ -1176,6 +1189,10 @@ function ensureCodexHooksJsonSessionStart(targetDir: string, opts: EnsureCodexSe
return { changed: false, wrote: false, path: hooksJsonPath };
}
managedCommand = shimIR.hookCommand;
configuredEntrypoints.push(
{ runtime: 'codex', configPath: hooksJsonPath, scriptPath: shimIR.cmdPath, platform, selfExecutable: true },
{ runtime: 'codex', configPath: hooksJsonPath, scriptPath, interpreterCandidates: [parseAbsoluteRunnerToken(absoluteRunner)], platform },
);
} else {
managedCommand = projectManagedHookCommand({
absoluteRunner,
@@ -1183,15 +1200,23 @@ function ensureCodexHooksJsonSessionStart(targetDir: string, opts: EnsureCodexSe
runtime: 'codex',
platform,
}) ?? undefined;
if (managedCommand) {
configuredEntrypoints.push({
runtime: 'codex',
configPath: hooksJsonPath,
scriptPath,
interpreterCandidates: [parseAbsoluteRunnerToken(absoluteRunner)],
platform,
});
}
}
if (!managedCommand) return { changed: false, wrote: false, path: hooksJsonPath };
const commandWindows = platform === 'win32'
? JSON.stringify(shellCmdProjection.posixNormalize(cmdShimPath))
: undefined;
return reconcileCodexHooksJsonSessionStart(targetDir, { managedCommand, commandWindows });
const result = reconcileCodexHooksJsonSessionStart(targetDir, { managedCommand, commandWindows });
return { ...result, configuredEntrypoints };
}
// ---------------------------------------------------------------------------
@@ -1397,7 +1422,77 @@ interface BuildHookCommandOpts {
runtime?: string;
hookShell?: string;
env?: NodeJS.ProcessEnv;
execPath?: string;
existsSync?: (p: string) => boolean;
configPath?: string;
configuredEntrypoints?: ConfiguredEntrypoint[];
}
function configuredEntrypointsForHook(
configDir: string,
hookName: string,
opts: BuildHookCommandOpts,
): ConfiguredEntrypoint[] {
const platform = opts.platform || process.platform;
const runtime = opts.runtime || 'generic';
const configPath = opts.configPath || configDir;
const target: ConfiguredEntrypoint = {
runtime,
configPath,
scriptPath: path.join(configDir, 'hooks', hookName),
platform,
};
const isShellHook = hookName.endsWith('.sh');
if (shellHookOmitsBashRunner({ platform, runtime, isShellHook })) return [{ ...target, selfExecutable: true }];
const bash = resolveBashExecutable(opts);
if (isShellHook) {
// An unresolved bash must still surface as an interpreterCandidates entry
// (the literal token, same as the portableHooks runner below) so
// validateConfiguredEntrypoints reports 'unresolved-interpreter' instead
// of silently skipping the check because the field is absent.
return [{ ...target, interpreterCandidates: [bash === null ? 'bash' : bash] }];
}
// #4249: check the SAME stable alias buildNodeRunnerChainToken bakes as its
// first candidate (normalizeNodePath rewrites a version-manager shim like
// fnm/nvm/mise/volta into its persistent path), not the raw, currently-
// running process.execPath — which always trivially resolves regardless of
// whether the alias actually baked into the persisted command still does.
const nodeCandidates = [
normalizeNodePath(opts.execPath || process.execPath, opts),
'node',
'/usr/local/bin/node',
'/usr/bin/node',
].filter((candidate): candidate is string => Boolean(candidate));
if (!opts.portableHooks) {
return [{ ...target, interpreterCandidates: nodeCandidates }];
}
const runner: ConfiguredEntrypoint = {
runtime,
configPath,
scriptPath: path.join(configDir, 'hooks', NODE_RUNNER_RESOLVER_HOOK),
interpreterCandidates: bash === null ? ['bash'] : [bash],
platform,
};
return [runner, { ...target, interpreterCandidates: nodeCandidates }];
}
function recordConfiguredHookCommand(
command: string | null,
configDir: string,
hookName: string,
opts: BuildHookCommandOpts,
): string | null {
if (command && opts.configuredEntrypoints) {
opts.configuredEntrypoints.push(
...configuredEntrypointsForHook(configDir, hookName, opts).map(entry => ({ ...entry, command })),
);
}
return command;
}
function buildHookCommand(configDir: string, hookName: string, opts?: BuildHookCommandOpts): string | null {
@@ -1406,6 +1501,8 @@ function buildHookCommand(configDir: string, hookName: string, opts?: BuildHookC
const runtime = opts.runtime || 'generic';
const hookShell = opts.hookShell;
const isShellHook = hookName.endsWith('.sh');
const track = (command: string | null): string | null =>
recordConfiguredHookCommand(command, configDir, hookName, opts);
if (shellHookOmitsBashRunner({ platform, runtime, isShellHook })) {
if (opts.portableHooks) {
@@ -1413,9 +1510,9 @@ function buildHookCommand(configDir: string, hookName: string, opts?: BuildHookC
configDir,
homeDir: os.homedir(),
});
return JSON.stringify(`${portableBaseDir}/hooks/${hookName}`);
return track(JSON.stringify(`${portableBaseDir}/hooks/${hookName}`));
}
return JSON.stringify(shellCmdProjection.posixNormalize(configDir) + '/hooks/' + hookName);
return track(JSON.stringify(shellCmdProjection.posixNormalize(configDir) + '/hooks/' + hookName));
}
// .sh hooks keep the pre-#3662 shape everywhere: the bash runner resolves
@@ -1423,30 +1520,42 @@ function buildHookCommand(configDir: string, hookName: string, opts?: BuildHookC
// (the absolute Git-Bash discovery covers win32 — #580/#3393).
if (isShellHook) {
const runner = resolveBashRunner(opts);
if (runner === null) return null;
if (runner === null) {
// #4249 (antigravity review): this early return skips `track()` below,
// so an unresolved bash on win32 (no Git Bash found) previously left
// this hook silently unregistered with nothing for
// validateConfiguredEntrypoints to reject — configuredEntrypointsForHook's
// own 'unresolved bash must still surface' comment describes intent this
// return never reached. Push the entry directly (no `command`, since
// none was ever built) so the gate actually sees it.
if (opts.configuredEntrypoints) {
opts.configuredEntrypoints.push(...configuredEntrypointsForHook(configDir, hookName, opts));
}
return null;
}
if (opts.portableHooks) {
const portableBaseDir = projectPortableHookBaseDir({
configDir,
homeDir: os.homedir(),
});
return projectManagedHookCommand({
return track(projectManagedHookCommand({
absoluteRunner: runner,
scriptPath: `${portableBaseDir}/hooks/${hookName}`,
runtime: opts.runtime || 'generic',
platform,
hookShell,
});
}));
}
const hooksPath = shellCmdProjection.posixNormalize(configDir) + '/hooks/' + hookName;
return projectManagedHookCommand({
return track(projectManagedHookCommand({
absoluteRunner: runner,
scriptPath: hooksPath,
runtime,
platform,
hookShell,
});
}));
}
// JS hooks (#3662): the node runner is resolved at hook-fire time, never
@@ -1470,7 +1579,7 @@ function buildHookCommand(configDir: string, hookName: string, opts?: BuildHookC
// Absolute Git-Bash discovery on win32 when available (#580); `bash` on
// PATH otherwise — the same assumption .sh hooks already make.
const resolverRunner = resolveBashRunner(opts) || 'bash';
return shellCmdProjection.projectShellCommandText({
return track(shellCmdProjection.projectShellCommandText({
runnerToken: resolverRunner,
argTokens: [
JSON.stringify(`${portableBaseDir}/hooks/${NODE_RUNNER_RESOLVER_HOOK}`),
@@ -1480,19 +1589,19 @@ function buildHookCommand(configDir: string, hookName: string, opts?: BuildHookC
runtime,
platform,
hookShell,
});
}));
}
const chainRunner = buildNodeRunnerChainToken(opts);
if (chainRunner === null) return null;
const hooksPath = shellCmdProjection.posixNormalize(configDir) + '/hooks/' + hookName;
return shellCmdProjection.projectShellCommandText({
return track(shellCmdProjection.projectShellCommandText({
runnerToken: chainRunner,
argTokens: [JSON.stringify(hooksPath)],
runtime,
platform,
hookShell,
});
}));
}
// ---------------------------------------------------------------------------
@@ -1600,8 +1709,9 @@ function mergeGsdAgentsMd(filePath: string, gsdContent: string): void {
// writeClineArtifacts
// ---------------------------------------------------------------------------
function writeClineArtifacts(targetDir: string, isGlobalInstall: boolean): string[] {
function writeClineArtifacts(targetDir: string, isGlobalInstall: boolean): { written: string[]; configuredEntrypoints: ConfiguredEntrypoint[] } {
const written: string[] = [];
const configuredEntrypoints: ConfiguredEntrypoint[] = [];
const clinerulesDir = path.join(targetDir, '.clinerules');
try {
@@ -1626,6 +1736,20 @@ function writeClineArtifacts(targetDir: string, isGlobalInstall: boolean): strin
try { fs.chmodSync(hookPath, 0o755); } catch { /* Windows: hooks unsupported anyway */ }
written.push('.clinerules/hooks/PreToolUse');
console.log(` ${green}✓${reset} Wrote .clinerules/hooks/PreToolUse`);
// #4249 (CodeRabbit): Cline invokes this file directly via its own
// `#!/usr/bin/env node` shebang — a hybrid case. The script itself still
// needs the execute bit (selfExecutable), but unlike GSD's other JS hooks
// (which bake an absolute, install-time-resolved node path specifically to
// avoid this) its interpreter is looked up on PATH by `env` at hook-fire
// time, so `node` must also resolve or the hook can never run.
configuredEntrypoints.push({
runtime: 'cline',
configPath: hookPath,
scriptPath: hookPath,
interpreterCandidates: ['node'],
selfExecutable: true,
platform: process.platform,
});
if (isGlobalInstall) {
try {
@@ -1637,7 +1761,7 @@ function writeClineArtifacts(targetDir: string, isGlobalInstall: boolean): strin
}
}
return written;
return { written, configuredEntrypoints };
}
// ---------------------------------------------------------------------------
@@ -1856,7 +1980,7 @@ function stageTransitiveHookLibs(opts: {
return staged;
}
function writeCursorHooksJson(targetDir: string, src: string, opts?: WriteCursorHooksJsonOpts): { hooksJsonPath: string; changed: boolean } {
function writeCursorHooksJson(targetDir: string, src: string, opts?: WriteCursorHooksJsonOpts): { hooksJsonPath: string; changed: boolean; configuredEntrypoints: ConfiguredEntrypoint[] } {
opts = opts || {};
const hooksDir = path.join(targetDir, 'hooks');
fs.mkdirSync(hooksDir, { recursive: true });
@@ -1915,7 +2039,13 @@ function writeCursorHooksJson(targetDir: string, src: string, opts?: WriteCursor
ensureCommonJsMarker(hooksDir);
}
const hookOpts: BuildHookCommandOpts = { runtime: 'cursor', platform: opts.platform || process.platform };
const configuredEntrypoints: ConfiguredEntrypoint[] = [];
const hookOpts: BuildHookCommandOpts = {
runtime: 'cursor',
platform: opts.platform || process.platform,
configPath: path.join(targetDir, 'hooks.json'),
configuredEntrypoints,
};
const commands: Record<string, string | null> = {};
for (const ev of events) {
const script = CURSOR_EVENT_SCRIPT_MAP[ev];
@@ -1929,7 +2059,7 @@ function writeCursorHooksJson(targetDir: string, src: string, opts?: WriteCursor
const hooksJsonPath = path.join(targetDir, 'hooks.json');
const result = reconcileCursorHooksJson(hooksJsonPath, managedEntries);
return { hooksJsonPath, changed: result.changed };
return { hooksJsonPath, changed: result.changed, configuredEntrypoints };
}
function removeCursorHooksJson(targetDir: string): { changed: boolean } {
@@ -2102,7 +2232,7 @@ interface WriteWindsurfHooksJsonOpts {
* @param opts - `{ platform? }`
* @returns `{ hooksJsonPath, changed }`
*/
function writeWindsurfHooksJson(targetDir: string, src: string, opts?: WriteWindsurfHooksJsonOpts): { hooksJsonPath: string; changed: boolean } {
function writeWindsurfHooksJson(targetDir: string, src: string, opts?: WriteWindsurfHooksJsonOpts): { hooksJsonPath: string; changed: boolean; configuredEntrypoints: ConfiguredEntrypoint[] } {
opts = opts || {};
const hooksDir = path.join(targetDir, 'hooks');
fs.mkdirSync(hooksDir, { recursive: true });
@@ -2154,7 +2284,13 @@ function writeWindsurfHooksJson(targetDir: string, src: string, opts?: WriteWind
ensureCommonJsMarker(hooksDir);
}
const hookOpts: BuildHookCommandOpts = { runtime: 'windsurf', platform: opts.platform || process.platform };
const configuredEntrypoints: ConfiguredEntrypoint[] = [];
const hookOpts: BuildHookCommandOpts = {
runtime: 'windsurf',
platform: opts.platform || process.platform,
configPath: path.join(targetDir, 'hooks.json'),
configuredEntrypoints,
};
const commands: Record<string, string | null> = {};
for (const ev of WINDSURF_HOOK_EVENTS) {
const script = WINDSURF_EVENT_SCRIPT_MAP[ev];
@@ -2169,7 +2305,7 @@ function writeWindsurfHooksJson(targetDir: string, src: string, opts?: WriteWind
const hooksJsonPath = path.join(targetDir, 'hooks.json');
const result = reconcileWindsurfHooksJson(hooksJsonPath, managedEntries);
return { hooksJsonPath, changed: result.changed };
return { hooksJsonPath, changed: result.changed, configuredEntrypoints };
}
/**
@@ -3150,30 +3286,38 @@ function writeKimiHooksToml(
configPath: string,
targetDir: string,
opts: { hookOpts: BuildHookCommandOpts },
): { changed: boolean; path: string; entryCount: number } {
): { changed: boolean; path: string; entryCount: number; configuredEntrypoints: ConfiguredEntrypoint[] } {
const configuredEntrypoints: ConfiguredEntrypoint[] = [];
const trackedOpts = {
hookOpts: {
...opts.hookOpts,
configPath,
configuredEntrypoints,
},
};
const existing = fs.existsSync(configPath) ? fs.readFileSync(configPath, 'utf8') : '';
const stripped = stripKimiHooksTomlBlock(existing) ?? '';
const block = buildKimiHooksTomlBlock(targetDir, opts);
const block = buildKimiHooksTomlBlock(targetDir, trackedOpts);
const entryCount = block ? (block.match(/\[\[hooks\]\]/g) || []).length : 0;
if (!block) {
if (stripped === existing) return { changed: false, path: configPath, entryCount: 0 };
if (stripped === existing) return { changed: false, path: configPath, entryCount: 0, configuredEntrypoints };
if (stripped.trim() === '') {
if (fs.existsSync(configPath)) fs.unlinkSync(configPath);
} else {
fs.mkdirSync(path.dirname(configPath), { recursive: true });
atomicWriteFileSync(configPath, stripped, 'utf8');
}
return { changed: true, path: configPath, entryCount: 0 };
return { changed: true, path: configPath, entryCount: 0, configuredEntrypoints };
}
const separator = stripped.trim() === '' ? '' : (stripped.endsWith('\n') ? '\n' : '\n\n');
const next = stripped.trim() === '' ? `${block}\n` : `${stripped}${separator}${block}\n`;
if (next === existing) return { changed: false, path: configPath, entryCount };
if (next === existing) return { changed: false, path: configPath, entryCount, configuredEntrypoints };
fs.mkdirSync(path.dirname(configPath), { recursive: true });
atomicWriteFileSync(configPath, next, 'utf8');
return { changed: true, path: configPath, entryCount };
return { changed: true, path: configPath, entryCount, configuredEntrypoints };
}
/**
@@ -3220,6 +3364,101 @@ function referencesHook(h: Record<string, unknown>, hookName: string): boolean {
(typeof url === 'string' && url.includes(hookName));
}
type ConfiguredEntrypointFailureReason = 'missing' | 'unreadable' | 'wrong-file-type' | 'unresolved-interpreter' | 'not-executable';
interface ConfiguredEntrypoint {
runtime: string;
configPath: string;
scriptPath: string;
interpreterCandidates?: string[];
// #4249 (CodeRabbit): true when the OS execs scriptPath directly (via a
// shebang, or Windows' own .cmd extension dispatch) — orthogonal to
// interpreterCandidates, which every producer that needs both sets
// alongside this rather than relying on their absence. Most self-executable
// entries have no candidates (a Windows-Claude .sh hook, Codex's .cmd shim);
// Cline's `#!/usr/bin/env node` is a hybrid needing both: the execute bit
// AND `node` resolving on PATH.
selfExecutable?: boolean;
platform?: string;
command?: string;
}
interface ConfiguredEntrypointInvalid {
runtime: string;
configPath: string;
role: 'script' | 'interpreter';
path: string;
reason: ConfiguredEntrypointFailureReason;
}
type ConfiguredEntrypointValidationResult =
| { ok: true }
| { ok: false; invalid: ConfiguredEntrypointInvalid[] };
function validateConfiguredEntrypoints(
entries: ConfiguredEntrypoint[],
deps: { statSync?: typeof fs.statSync; accessSync?: typeof fs.accessSync; resolveExecutableBinary?: typeof resolveExecutableBinary } = {},
): ConfiguredEntrypointValidationResult {
const statSync = deps.statSync ?? fs.statSync;
const accessSync = deps.accessSync ?? fs.accessSync;
const resolve = deps.resolveExecutableBinary ?? resolveExecutableBinary;
const invalid: ConfiguredEntrypointInvalid[] = [];
for (const entry of entries) {
let scriptOk = false;
try {
scriptOk = statSync(entry.scriptPath).isFile();
if (!scriptOk) {
invalid.push({ runtime: entry.runtime, configPath: entry.configPath, role: 'script', path: entry.scriptPath, reason: 'wrong-file-type' });
} else {
// #4249: statSync only needs search permission on the parent dirs, so
// it succeeds even for a chmod-000 file — the EACCES catch below
// never fires for that case. Read permission on the file itself must
// be checked explicitly: an interpreter opens the script directly,
// and even a self-executable shebang script is opened and read by
// its kernel-invoked interpreter, not just exec'd — X_OK alone does
// not prove it's readable.
try {
accessSync(entry.scriptPath, fs.constants.R_OK);
} catch {
scriptOk = false;
invalid.push({ runtime: entry.runtime, configPath: entry.configPath, role: 'script', path: entry.scriptPath, reason: 'unreadable' });
}
}
} catch (statErr) {
// #4249 Nit: EACCES means a parent directory couldn't be searched —
// a real (if rare) permission problem, distinct from ENOENT's "missing".
// EPERM: Windows' equivalent permission-denied code for a directory a
// parent path couldn't be traversed into.
const code = (statErr as NodeJS.ErrnoException)?.code;
const reason = code === 'EACCES' || code === 'EPERM' ? 'unreadable' : 'missing';
invalid.push({ runtime: entry.runtime, configPath: entry.configPath, role: 'script', path: entry.scriptPath, reason });
}
// #4249: selfExecutable is the sole source of truth for whether the OS
// execs scriptPath directly via its own shebang (set explicitly by every
// producer that needs it — a Windows-Claude .sh hook, Codex's Windows
// .cmd shim, Cline's hybrid `env node` hook — rather than inferred from
// the absence of interpreterCandidates, which Cline's hybrid case also
// carries). Skip on win32 like resolveExecutableBinary's own X_OK
// carve-out does: POSIX mode bits don't mean executable on Windows, and a
// real accessSync(X_OK) there would fail a .cmd shim under a test that
// simulates win32 on a POSIX runner (Node's own no-op only protects an
// actual Windows machine). Cline is the only producer where this runs.
if (scriptOk && entry.selfExecutable && (entry.platform ?? process.platform) !== 'win32') {
try {
accessSync(entry.scriptPath, fs.constants.X_OK);
} catch {
invalid.push({ runtime: entry.runtime, configPath: entry.configPath, role: 'script', path: entry.scriptPath, reason: 'not-executable' });
}
}
if (entry.interpreterCandidates && !entry.interpreterCandidates.some(candidate =>
resolve(candidate, { platform: entry.platform, requireExecutable: true }) !== null,
)) {
invalid.push({ runtime: entry.runtime, configPath: entry.configPath, role: 'interpreter', path: entry.interpreterCandidates.join(' | '), reason: 'unresolved-interpreter' });
}
}
return invalid.length === 0 ? { ok: true } : { ok: false, invalid };
}
// ---------------------------------------------------------------------------
// Exports
// ---------------------------------------------------------------------------
@@ -3295,7 +3534,9 @@ export = {
// Shared
stageTransitiveHookLibs,
buildHookCommand,
recordConfiguredHookCommand,
applySettingsJsonHooks,
validateConfiguredEntrypoints,
referencesHook,
rewriteLegacyManagedNodeHookCommands,
reconcileManagedShellHookCommands,

View File

@@ -249,6 +249,7 @@ const MANAGED_HOOK_COMMAND_BASENAMES_BY_SURFACE: Record<string, Set<string>> = {
'gsd-session-state.sh',
'gsd-validate-commit.sh',
'gsd-phase-boundary.sh',
'gsd-graphify-update.sh',
// #3662: same three guards as MANAGED_HOOK_BASENAMES_BY_SURFACE above —
// their absence here meant isManagedHookCommand never recognized them, so
// the settings.json→settings.local.json migration filter (and uninstall

View File

@@ -2361,6 +2361,95 @@ describe('#3245 — idempotent rollback reverts skills/, agents/, and VERSION',
});
}
// ────────────────────────────────────────────────────────────────────────
// #4249 — install() exposes the full snapshot restore, not just migrations rollback
// ────────────────────────────────────────────────────────────────────────
{
const { test, describe, beforeEach, afterEach } = require('node:test');
const os = require('os');
const { cleanup } = require('./helpers.cjs');
const previousGsdTestMode = process.env.GSD_TEST_MODE;
process.env.GSD_TEST_MODE = '1';
const { install } = require('../bin/install.js');
if (previousGsdTestMode === undefined) {
delete process.env.GSD_TEST_MODE;
} else {
process.env.GSD_TEST_MODE = previousGsdTestMode;
}
// concurrency: false — drives the real install pipeline like the block above.
describe('#4249 — install() exposes the full snapshot restore, not just migrations rollback', { concurrency: false }, () => {
let tmpDir;
let codexHome;
beforeEach(() => {
tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), 'gsd-4249-codex-rollback-'));
codexHome = path.join(tmpDir, 'codex-home');
fs.mkdirSync(codexHome, { recursive: true });
});
afterEach(() => cleanup(tmpDir));
test('result.rollbackPreInstallSnapshot() reverts skills/, agents/, and VERSION', () => {
// Skills resolve $HOME-relative independent of CODEX_HOME (#2088), so the
// rollback closure must run before this sandboxing is torn down — inline
// runCodexInstall's env dance instead of using the auto-restoring helper,
// matching how installAllRuntimes' real aggregate gate calls it: in the
// same process env install() itself ran in, never after it's restored.
const previousHome = process.env.HOME;
const previousUserProfile = process.env.USERPROFILE;
const previousCodexHome = process.env.CODEX_HOME;
const previousCwd = process.cwd();
process.env.HOME = codexHome;
process.env.USERPROFILE = codexHome;
process.env.CODEX_HOME = codexHome;
process.chdir(path.join(__dirname, '..'));
try {
const result = install(true, 'codex');
// A configured-entrypoint validation failure discovered outside install()
// (installAllRuntimes' aggregate assertConfiguredEntrypoints, run after
// this function already returned) reaches for this field. Before #4249
// the only rollback Codex exposed was rollbackInstallerMigrations, which
// reverts installer-migration state only and leaves the skills/agents/
// VERSION this successful install just wrote untouched.
result.rollbackPreInstallSnapshot();
const skillsDir = codexSkillsRoot(codexHome);
const gsdSkills = fs.existsSync(skillsDir)
? fs.readdirSync(skillsDir, { withFileTypes: true }).filter(e => e.isDirectory() && e.name.startsWith('gsd-'))
: [];
assert.strictEqual(gsdSkills.length, 0, 'rollback must remove all gsd-* skill directories: ' + gsdSkills.map(e => e.name).join(', '));
const versionPath = path.join(codexHome, 'gsd-core', 'VERSION');
assert.strictEqual(fs.existsSync(versionPath), false, 'rollback must remove gsd-core/VERSION');
// #4249 (agy adversarial review): the whole point of this describe block
// is that rollback covers the full pre-install snapshot, not just
// installer migrations — config.toml/hooks.json must revert too. Both
// were absent before this fresh install, so rollback must remove them.
assert.strictEqual(
fs.existsSync(path.join(codexHome, 'config.toml')),
false,
'rollback must remove config.toml (absent before this fresh install)'
);
assert.strictEqual(
fs.existsSync(path.join(codexHome, 'hooks.json')),
false,
'rollback must remove hooks.json (absent before this fresh install)'
);
} finally {
process.chdir(previousCwd);
if (previousHome === undefined) delete process.env.HOME;
else process.env.HOME = previousHome;
if (previousUserProfile === undefined) delete process.env.USERPROFILE;
else process.env.USERPROFILE = previousUserProfile;
if (previousCodexHome === undefined) delete process.env.CODEX_HOME;
else process.env.CODEX_HOME = previousCodexHome;
}
});
});
}
// ────────────────────────────────────────────────────────────────────────
// Folded from tests/bug-3285-codex-hooks-state-allowed.test.cjs — consolidation epic #1969 (B1 #1970)

View File

@@ -0,0 +1,558 @@
'use strict';
const assert = require('node:assert/strict');
const fs = require('node:fs');
const os = require('node:os');
const path = require('node:path');
const test = require('node:test');
const helpers = require('./helpers.cjs');
const hooksSurface = require('../gsd-core/bin/lib/runtime-hooks-surface.cjs');
const { install, installAllRuntimes, finishInstall } = require('../bin/install.js');
/**
* Run `fn` with HOME/USERPROFILE pointed at a fresh temp dir and every
* config-location env var scrubbed, so a real install() never touches the
* developer's own config. Restores all of it, and cleans the dir, afterwards.
*/
function withSandboxedHome(t, prefix, fn) {
const root = fs.mkdtempSync(path.join(os.tmpdir(), prefix));
t.after(() => helpers.cleanup(root));
const savedHome = process.env.HOME;
const savedUserProfile = process.env.USERPROFILE;
process.env.HOME = root;
process.env.USERPROFILE = root;
const restoreConfigLocationEnv = helpers.scrubConfigLocationEnv();
try {
return fn(root);
} finally {
if (savedHome === undefined) delete process.env.HOME;
else process.env.HOME = savedHome;
if (savedUserProfile === undefined) delete process.env.USERPROFILE;
else process.env.USERPROFILE = savedUserProfile;
restoreConfigLocationEnv();
}
}
test('configured entrypoint validation exposes an aggregate typed boundary', () => {
assert.equal(
typeof hooksSurface.validateConfiguredEntrypoints,
'function',
'the Runtime Hooks Surface must export configured-entrypoint validation',
);
});
test('finishInstall rejects an invalid configured entrypoint before Done output', (t) => {
const root = fs.mkdtempSync(path.join(os.tmpdir(), 'configured-entrypoint-finish-'));
t.after(() => helpers.cleanup(root));
const logs = [];
const originalLog = console.log;
console.log = (...args) => logs.push(args.join(' '));
// #2665/#4249: finishInstall asserts configured entrypoints before any of its
// own writes now, but still sandbox HOME (+ USERPROFILE for os.homedir() on
// Windows) and config-location env defensively, so a future reordering that
// reintroduces a pre-assertion write can never redirect it to a live config dir.
const savedHome = process.env.HOME;
const savedUserProfile = process.env.USERPROFILE;
process.env.HOME = root;
process.env.USERPROFILE = root;
const restoreConfigLocationEnv = helpers.scrubConfigLocationEnv();
try {
assert.throws(() => finishInstall(null, null, null, false, 'cline', false, root, {
configuredEntrypoints: [{ runtime: 'cline', configPath: path.join(root, 'config'), scriptPath: path.join(root, 'missing.js') }],
}), /Configured entrypoint validation failed/);
// #4249 review, Major: the message must name the actual consequence for
// the failing runtime, not just that something failed — Cline has no
// revert path, so its entry must be flagged "NOT reverted".
assert.throws(() => finishInstall(null, null, null, false, 'cline', false, root, {
configuredEntrypoints: [{ runtime: 'cline', configPath: path.join(root, 'config'), scriptPath: path.join(root, 'missing.js') }],
}), /NOT reverted/);
} finally {
console.log = originalLog;
restoreConfigLocationEnv();
if (savedHome === undefined) delete process.env.HOME;
else process.env.HOME = savedHome;
if (savedUserProfile === undefined) delete process.env.USERPROFILE;
else process.env.USERPROFILE = savedUserProfile;
}
assert.equal(logs.some(line => line.includes('Done!')), false);
});
test('a hook already registered under a stale command is still tracked for validation on re-install (#4154 Blocker)', (t) => {
withSandboxedHome(t, 'configured-entrypoint-stale-', () => {
const first = install(true, 'claude');
assert.ok(first.settingsPath, 'a fresh global install must produce a settings path');
finishInstall(first.settingsPath, first.settings, first.statuslineCommand, false, 'claude', true, first.configDir, {
configuredEntrypoints: first.configuredEntrypoints,
});
// Simulate an entry registered by an older installer under a DIFFERENT
// node install (e.g. an nvm switch, #4087/#4098/#4137): same real
// scriptPath under <configDir>/hooks/ (that never changes across
// installer versions) and still shaped as the modern runtime-resolving
// chain (rewriteLegacyManagedNodeHookCommands deliberately never touches
// an already-current-format entry — #3662), but baked with a node path
// this install would never produce. `hasGsdUpdateHook` still finds it and
// applySettingsJsonHooks takes its register-only-if-absent branch on the
// next install (never rewriting it).
const onDisk = JSON.parse(fs.readFileSync(first.settingsPath, 'utf8'));
const staleEntry = (onDisk.hooks.SessionStart || []).find(entry =>
entry.hooks && entry.hooks.some(h => h.command && h.command.includes('gsd-check-update.js'))
);
assert.ok(staleEntry, 'a fresh install must register the check-update hook');
const staleCommand = hooksSurface.buildHookCommand(first.configDir, 'gsd-check-update.js', {
execPath: '/old/nvm/pinned/node',
platform: process.platform,
runtime: 'claude',
});
for (const h of staleEntry.hooks) {
if (h.command && h.command.includes('gsd-check-update.js')) {
h.command = staleCommand;
}
}
fs.writeFileSync(first.settingsPath, JSON.stringify(onDisk, null, 2));
const second = install(true, 'claude');
const trackedNames = (second.configuredEntrypoints || []).map(entry => path.basename(entry.scriptPath));
assert.ok(
trackedNames.includes('gsd-check-update.js'),
`a hook already registered under a stale command must still be tracked for validation, got: ${trackedNames.join(', ')}`,
);
});
});
test('configured entrypoint validation aggregates file and interpreter failures without execution', (t) => {
const root = fs.mkdtempSync(path.join(os.tmpdir(), 'configured-entrypoint-'));
t.after(() => helpers.cleanup(root));
const directory = path.join(root, 'directory');
fs.mkdirSync(directory);
const unreadablePath = path.join(root, 'unreadable.js');
const notExecutablePath = path.join(root, 'not-executable.js');
const result = hooksSurface.validateConfiguredEntrypoints([
{ runtime: 'claude', configPath: path.join(root, 'settings.json'), scriptPath: path.join(root, 'missing.js') },
{ runtime: 'claude', configPath: path.join(root, 'settings.json'), scriptPath: directory },
{ runtime: 'claude', configPath: path.join(root, 'settings.json'), scriptPath: __filename, interpreterCandidates: ['missing-node'] },
{ runtime: 'claude', configPath: path.join(root, 'settings.json'), scriptPath: unreadablePath },
// #4249: selfExecutable means this entry is invoked directly via its own
// shebang (e.g. a Windows-Claude .sh hook) — must itself be +x.
// platform pinned to non-win32: the X_OK check itself is a POSIX-only
// concept (skipped entirely on win32, matching production) — this case
// must exercise it deterministically regardless of which OS runs the test.
{ runtime: 'claude', configPath: path.join(root, 'settings.json'), scriptPath: notExecutablePath, selfExecutable: true, platform: 'linux' },
], {
resolveExecutableBinary: () => null,
statSync: (p) => {
if (p === unreadablePath) {
const err = new Error('EACCES: permission denied');
err.code = 'EACCES';
throw err;
}
return fs.statSync(p === notExecutablePath ? __filename : p);
},
accessSync: (p, mode) => {
if (p === notExecutablePath && mode === fs.constants.X_OK) {
const err = new Error('EACCES: permission denied');
err.code = 'EACCES';
throw err;
}
// notExecutablePath is never written to disk (its statSync mock above
// redirects to a real file instead) — its R_OK call must redirect too,
// or this falls through to a real accessSync on a nonexistent path.
return fs.accessSync(p === notExecutablePath ? __filename : p, mode);
},
});
assert.equal(result.ok, false);
assert.deepEqual(result.invalid.map(({ role, reason }) => [role, reason]), [
['script', 'missing'],
['script', 'wrong-file-type'],
['interpreter', 'unresolved-interpreter'],
['script', 'unreadable'],
['script', 'not-executable'],
]);
});
test('an interpreter-invoked script that exists but has no read permission is reported unreadable, not ok (#4249 agy review)', (t) => {
// statSync only needs search (+x) permission on the parent directories, so
// it succeeds on a chmod-000 file even though `node <script>` would fail
// with EACCES at hook-fire time.
const root = fs.mkdtempSync(path.join(os.tmpdir(), 'configured-entrypoint-unreadable-interp-'));
t.after(() => helpers.cleanup(root));
const scriptPath = path.join(root, 'hook.js');
fs.writeFileSync(scriptPath, '// unreadable to the invoking user at hook-fire time\n');
const result = hooksSurface.validateConfiguredEntrypoints([
{
runtime: 'claude',
configPath: path.join(root, 'settings.json'),
scriptPath,
interpreterCandidates: ['/usr/bin/node'],
},
], {
resolveExecutableBinary: () => '/usr/bin/node',
accessSync: (p, mode) => {
if (p === scriptPath && mode === fs.constants.R_OK) {
const err = new Error('EACCES: permission denied');
err.code = 'EACCES';
throw err;
}
return fs.accessSync(p, mode);
},
});
assert.equal(result.ok, false);
assert.deepEqual(result.invalid.map(({ role, reason }) => [role, reason]), [
['script', 'unreadable'],
]);
});
test('an EPERM from statSync (Windows equivalent of EACCES on a parent directory) is also reported unreadable, not missing (#4249 agy review)', (t) => {
const root = fs.mkdtempSync(path.join(os.tmpdir(), 'configured-entrypoint-eperm-'));
t.after(() => helpers.cleanup(root));
const scriptPath = path.join(root, 'hook.js');
const result = hooksSurface.validateConfiguredEntrypoints([
{ runtime: 'claude', configPath: path.join(root, 'settings.json'), scriptPath },
], {
statSync: () => {
const err = new Error('EPERM: operation not permitted');
err.code = 'EPERM';
throw err;
},
});
assert.equal(result.ok, false);
assert.deepEqual(result.invalid.map(({ role, reason }) => [role, reason]), [
['script', 'unreadable'],
]);
});
test('a selfExecutable + interpreterCandidates entry checks both the execute bit and the interpreter (#4249 CodeRabbit review — Cline\'s hybrid `env node` shebang)', (t) => {
const root = fs.mkdtempSync(path.join(os.tmpdir(), 'configured-entrypoint-self-exec-'));
t.after(() => helpers.cleanup(root));
const okScript = path.join(root, 'ok.js');
const notExecScript = path.join(root, 'not-exec.js');
fs.writeFileSync(okScript, '#!/usr/bin/env node\n');
fs.writeFileSync(notExecScript, '#!/usr/bin/env node\n');
const makeEntry = (scriptPath) => ({
runtime: 'cline',
configPath: path.join(root, '.clinerules', 'hooks', 'PreToolUse'),
scriptPath,
interpreterCandidates: ['node'],
selfExecutable: true,
// X_OK is a POSIX-only concept (skipped entirely on win32, matching
// production) — pinned here so the execute-bit case below is exercised
// deterministically regardless of which OS runs the test.
platform: 'linux',
});
// plain writeFileSync never sets the execute bit (and the repo bans chmod
// in tests), so every case below stubs accessSync to simulate the exact
// permission state under test rather than relying on the real filesystem.
const alwaysOk = () => {};
// node missing from PATH entirely: caught even though the script itself is fine.
const missingInterpreter = hooksSurface.validateConfiguredEntrypoints([makeEntry(okScript)], {
resolveExecutableBinary: () => null,
accessSync: alwaysOk,
});
assert.equal(missingInterpreter.ok, false);
assert.deepEqual(missingInterpreter.invalid.map(({ role, reason }) => [role, reason]), [
['interpreter', 'unresolved-interpreter'],
]);
// node resolves fine, but the script itself lost its execute bit: caught too —
// interpreterCandidates being present must not skip the X_OK check here.
const notExecutable = hooksSurface.validateConfiguredEntrypoints([makeEntry(notExecScript)], {
resolveExecutableBinary: () => '/usr/bin/node',
accessSync: (p, mode) => {
if (p === notExecScript && mode === fs.constants.X_OK) {
const err = new Error('EACCES: permission denied');
err.code = 'EACCES';
throw err;
}
},
});
assert.equal(notExecutable.ok, false);
assert.deepEqual(notExecutable.invalid.map(({ role, reason }) => [role, reason]), [
['script', 'not-executable'],
]);
// both hold: clean pass.
const clean = hooksSurface.validateConfiguredEntrypoints([makeEntry(okScript)], {
resolveExecutableBinary: () => '/usr/bin/node',
accessSync: alwaysOk,
});
assert.equal(clean.ok, true);
});
test('a win32 selfExecutable entry (Codex .cmd shim) skips the execute-bit check (#4249 agy review)', (t) => {
// POSIX mode bits don't mean executable on win32 — this must not depend on
// Node's own accessSync(X_OK)-as-F_OK no-op (which only protects a real
// Windows machine), or a test simulating win32 on a POSIX runner would see
// a spurious 'not-executable' for a .cmd shim that was never chmod'd +x.
const root = fs.mkdtempSync(path.join(os.tmpdir(), 'configured-entrypoint-win32-cmd-'));
t.after(() => helpers.cleanup(root));
const cmdPath = path.join(root, 'gsd-check-update.cmd');
fs.writeFileSync(cmdPath, '@echo off\r\n');
const result = hooksSurface.validateConfiguredEntrypoints([
{
runtime: 'codex',
configPath: path.join(root, 'hooks.json'),
scriptPath: cmdPath,
platform: 'win32',
selfExecutable: true,
},
], {
accessSync: (p, mode) => {
if (mode === fs.constants.X_OK) {
const err = new Error('EACCES: permission denied');
err.code = 'EACCES';
throw err;
}
},
});
assert.equal(result.ok, true, JSON.stringify(result));
});
test('runtime config writers expose the exact configured entrypoints they emit', (t) => {
const root = fs.mkdtempSync(path.join(os.tmpdir(), 'configured-entrypoint-writers-'));
t.after(() => helpers.cleanup(root));
const sourceRoot = path.join(__dirname, '..');
const codexRoot = path.join(root, 'codex');
// On win32, ensureCodexHooksJsonSessionStart writes a .cmd shim under
// <codexRoot>/hooks/; the real installer only calls this once that dir
// (and gsd-check-update.js) already exist, so create it here too.
fs.mkdirSync(path.join(codexRoot, 'hooks'), { recursive: true });
const codex = hooksSurface.ensureCodexHooksJsonSessionStart(codexRoot, {
absoluteRunner: JSON.stringify(process.execPath),
});
// win32 emits two entries (the .cmd shim plus the underlying script);
// every other platform emits the script alone — assert the shape common
// to both rather than a platform-fixed array.
assert.ok(codex.configuredEntrypoints.length >= 1);
assert.ok(codex.configuredEntrypoints.every(entry => entry.runtime === 'codex'
&& entry.configPath === path.join(codexRoot, 'hooks.json')
&& entry.platform === process.platform));
assert.ok(codex.configuredEntrypoints.some(
entry => path.basename(entry.scriptPath) === 'gsd-check-update.js'
&& Array.isArray(entry.interpreterCandidates)
&& entry.interpreterCandidates.length === 1,
));
const cursorRoot = path.join(root, 'cursor');
const cursor = hooksSurface.writeCursorHooksJson(cursorRoot, sourceRoot, {
managedHookEvents: ['sessionStart'],
});
assert.deepEqual(
cursor.configuredEntrypoints.map(entry => path.basename(entry.scriptPath)),
['gsd-cursor-session-start.js'],
);
assert.ok(cursor.configuredEntrypoints.every(entry => entry.configPath === cursor.hooksJsonPath));
const windsurfRoot = path.join(root, 'windsurf');
const windsurf = hooksSurface.writeWindsurfHooksJson(windsurfRoot, sourceRoot);
assert.deepEqual(
windsurf.configuredEntrypoints.map(entry => path.basename(entry.scriptPath)).sort(),
['gsd-windsurf-pre-command.js', 'gsd-windsurf-pre-write.js'],
);
const kimiRoot = path.join(root, 'kimi');
fs.cpSync(path.join(sourceRoot, 'hooks'), path.join(kimiRoot, 'hooks'), { recursive: true });
const kimiConfig = path.join(root, 'kimi-config.toml');
const kimi = hooksSurface.writeKimiHooksToml(kimiConfig, kimiRoot, {
hookOpts: { runtime: 'kimi' },
});
assert.equal(kimi.configuredEntrypoints.length, kimi.entryCount);
assert.ok(kimi.configuredEntrypoints.every(entry => entry.configPath === kimiConfig));
const clineRoot = path.join(root, 'cline');
const cline = hooksSurface.writeClineArtifacts(clineRoot, false);
assert.deepEqual(
cline.configuredEntrypoints.map(entry => path.basename(entry.scriptPath)),
['PreToolUse'],
);
// #4249 (CodeRabbit): Cline's `#!/usr/bin/env node` hook is a hybrid —
// self-executable (needs its own execute bit checked) AND PATH-dependent
// on `node` (needs an interpreter candidate resolved), unlike GSD's other
// JS hooks which bake an absolute node path specifically to avoid this.
assert.equal(cline.configuredEntrypoints[0].selfExecutable, true);
assert.deepEqual(cline.configuredEntrypoints[0].interpreterCandidates, ['node']);
const portable = [];
assert.ok(hooksSurface.buildHookCommand(root, 'gsd-context-monitor.js', {
runtime: 'claude',
portableHooks: true,
configPath: path.join(root, 'settings.json'),
configuredEntrypoints: portable,
}));
assert.deepEqual(
portable.map(entry => path.basename(entry.scriptPath)),
['gsd-node-runner.sh', 'gsd-context-monitor.js'],
);
});
test('a minimal Codex rollback restores skills from the alternate skills home (#4249 CodeRabbit)', (t) => {
withSandboxedHome(t, 'configured-entrypoint-codex-skills-', (root) => {
const first = install(true, 'codex');
// Codex resolves skills to $HOME/.agents/skills, NOT configDir (ADR-1239
// skills-kind `home` override) — the surface #3245's own snapshot owns and
// the manifest-driven pass deliberately leaves to it.
const skillsRoot = path.join(root, '.agents', 'skills');
const skillDir = fs.readdirSync(skillsRoot).find(name => name.startsWith('gsd-'));
assert.ok(skillDir, `a Codex install must write gsd-* skills under ${skillsRoot}`);
const skillFile = path.join(skillsRoot, skillDir, 'SKILL.md');
const priorBytes = '# bytes only this test wrote\n';
fs.writeFileSync(skillFile, priorBytes);
// Marker-driven core profile — the `gsd update` path into minimal mode.
fs.writeFileSync(path.join(first.configDir, '.gsd-profile'), 'core\n');
const second = install(true, 'codex');
second.rollbackPreInstallSnapshot();
assert.equal(
fs.existsSync(skillFile),
true,
'a minimal-mode rollback must not delete a skill dir it never snapshotted',
);
assert.equal(
fs.readFileSync(skillFile, 'utf8'),
priorBytes,
'a minimal-mode rollback must restore alternate-home skills byte-for-byte',
);
});
});
test('an aggregate entrypoint validation failure rolls the Codex install back (#4249)', (t) => {
withSandboxedHome(t, 'configured-entrypoint-aggregate-', () => {
const first = install(true, 'codex');
// config.toml is the surface #3245's snapshot restores and the one this
// runtime's writer rewrites on every install, so a sentinel surviving here
// can only be the rollback's doing.
const configPath = path.join(first.configDir, 'config.toml');
const priorBytes = '# bytes only this test wrote\n';
fs.writeFileSync(configPath, priorBytes);
// Cline's `#!/usr/bin/env node` hook records interpreterCandidates: ['node'],
// resolved off PATH. Emptying PATH makes exactly that entry fail, which is
// the only way to drive a REAL aggregate failure through installAllRuntimes:
// every other entrypoint is a path the installer just wrote. Codex installs
// first, so the reversed rollback loop reaches its restoreCodexSnapshot.
const savedPath = process.env.PATH;
process.env.PATH = '';
try {
assert.throws(
() => installAllRuntimes(['codex', 'cline'], true, false),
/Configured entrypoint validation failed/,
'an unresolvable interpreter must fail the aggregate gate',
);
} finally {
process.env.PATH = savedPath;
}
assert.equal(
fs.readFileSync(configPath, 'utf8'),
priorBytes,
'the aggregate failure must reach restoreCodexSnapshot, not just throw',
);
});
});
test('an aggregate entrypoint validation failure leaves Cline\'s own config file on disk, unreverted (#4249 review, Major)', (t) => {
withSandboxedHome(t, 'configured-entrypoint-unreverted-', () => {
// Cline is one of the four runtimes (Cursor/Windsurf/Kimi/Cline) that
// write their config file inside install(), ahead of the validation
// gate, with no snapshot/restore path — see the rollback-matrix
// paragraph in the update-gsd how-to. Unlike the Codex companion test above,
// this asserts the POSITIVE case that same paragraph discloses in
// prose but no prior test proved: the file the failing runtime itself
// just wrote is still there afterward, broken and unreverted.
const clineFirst = install(true, 'cline');
const clineHookPath = path.join(clineFirst.configDir, '.clinerules', 'hooks', 'PreToolUse');
assert.equal(fs.existsSync(clineHookPath), true, 'precondition: Cline hook must exist before the failing install');
// Same technique as the Codex rollback test: an emptied PATH makes
// Cline's own `#!/usr/bin/env node` hook fail interpreter resolution,
// driving a REAL aggregate failure through installAllRuntimes.
const savedPath = process.env.PATH;
process.env.PATH = '';
try {
assert.throws(
() => installAllRuntimes(['codex', 'cline'], true, false),
/Configured entrypoint validation failed/,
'an unresolvable interpreter must fail the aggregate gate',
);
} finally {
process.env.PATH = savedPath;
}
assert.equal(
fs.existsSync(clineHookPath),
true,
'Cline has no snapshot/restore path — its already-written hook file must remain on disk, not be deleted',
);
});
});
test('a NON-entrypoint finalize failure leaves an already-successful Codex install in place (#4249)', (t) => {
withSandboxedHome(t, 'configured-entrypoint-nonentry-', () => {
const first = install(true, 'codex');
// config.toml sits inside surface #3245's pre-install snapshot, so these
// bytes come back if — and only if — the full restoreCodexSnapshot() runs.
const configPath = path.join(first.configDir, 'config.toml');
const priorBytes = '# bytes only this test wrote\n';
fs.writeFileSync(configPath, priorBytes);
// Kilo is the smallest real non-Codex runtime whose finishInstall still
// writes: plan.finishPermissionWriter === 'kilo' makes it call
// configureKiloPermissions unconditionally (that writer, unlike OpenCode's,
// is NOT GSD_TEST_MODE-gated), ending in an unguarded
// fs.writeFileSync(<configDir>/kilo.json). Cline can't stand in here — its
// plan is writesSharedSettings:false + finishPermissionWriter:null, so its
// finishInstall performs no write at all and has no non-entrypoint failure
// path to force. EACCES on the Kilo write is injected by monkeypatching
// node:fs and restoring it in a finally — never chmod 0o000, which root
// bypasses under Docker/CI and would give this test zero real coverage.
const realWriteFileSync = fs.writeFileSync;
fs.writeFileSync = function poisonedWriteFileSync(target, ...args) {
const targetStr = typeof target === 'string' ? target : String(target);
if (/[\\/]kilo\.jsonc?$/.test(targetStr)) {
const injected = new Error(`EACCES: permission denied, open '${targetStr}'`);
injected.code = 'EACCES';
throw injected;
}
return realWriteFileSync.apply(fs, [target, ...args]);
};
try {
assert.throws(
() => installAllRuntimes(['codex', 'kilo'], true, false),
/EACCES: permission denied/,
'a failed Kilo permission write must abort the aggregate install',
);
} finally {
fs.writeFileSync = realWriteFileSync;
}
// Codex installs first and finishInstall prints its "Done!" summary before
// Kilo's throws, so the reversed rollback loop does reach Codex's entry —
// but with a non-entrypoint error it must take the installer-migrations-only
// closure, not restoreCodexSnapshot. A configured-entrypoint validation
// failure is the only trigger docs/how-to/update-gsd.md documents for
// reverting a runtime install that otherwise succeeded; un-installing (on
// update, downgrading) Codex because an unrelated runtime hit EACCES would
// contradict the "Done!" the user has already been shown.
assert.notEqual(
fs.readFileSync(configPath, 'utf8'),
priorBytes,
'a non-entrypoint finalize failure must not restore the Codex pre-install snapshot',
);
});
});

View File

@@ -562,13 +562,16 @@ describe('install/uninstall — trae (nested skills/gsd-<router>/skills/<stem>/
const result = install(false, 'trae');
const targetDir = path.join(tmpDir, '.trae');
assert.deepStrictEqual(result, {
assert.equal(typeof result.rollbackInstallerMigrations, 'function');
const { rollbackInstallerMigrations: _rollbackInstallerMigrations, ...resultWithoutRollback } = result;
assert.deepStrictEqual(resultWithoutRollback, {
settingsPath: null,
settings: null,
statuslineCommand: null,
updateBannerCommand: null,
runtime: 'trae',
configDir: fs.realpathSync(targetDir),
configuredEntrypoints: [],
});
// trae nests: skills/gsd-<router>/skills/<stem>/SKILL.md
@@ -5466,6 +5469,36 @@ describe('#338 case 1: fresh local Claude install writes to settings.local.json'
`install() must return settingsPath ending in settings.local.json for local Claude installs; got: ${result.settingsPath}`
);
});
// #4249 (agy adversarial review, round 8): every early-return branch of
// install() must still return the configuredEntrypoints/rollbackInstallerMigrations
// shape — rollbackFinalizedInstallerMigrations reads the latter unconditionally,
// so an omission here silently drops this runtime's rollback on a later
// finalize-stage failure.
test('malformed settings.local.json still returns configuredEntrypoints and a working rollbackInstallerMigrations', (t) => {
const origCwd = process.cwd();
t.after(() => { process.chdir(origCwd); });
process.chdir(tmpDir);
fs.mkdirSync(path.join(tmpDir, '.claude'), { recursive: true });
fs.writeFileSync(path.join(tmpDir, '.claude', 'settings.local.json'), '{ not valid json', 'utf-8');
const result = install(false, 'claude');
assert.deepStrictEqual(
result.configuredEntrypoints,
[],
'unparseable settings.local.json must still return configuredEntrypoints: []'
);
assert.strictEqual(
typeof result.rollbackInstallerMigrations,
'function',
'unparseable settings.local.json must still return a callable rollbackInstallerMigrations'
);
assert.doesNotThrow(
() => result.rollbackInstallerMigrations(),
'rollbackInstallerMigrations() must not throw when called on this early-return path'
);
});
});
// ─── Case 2: global Claude install (regression guard) ────────────────────────

View File

@@ -10,12 +10,14 @@ process.env.GSD_TEST_MODE = '1';
const { describe, test, before, after } = require('node:test');
const assert = require('node:assert/strict');
const fs = require('node:fs');
const os = require('node:os');
const path = require('node:path');
const crypto = require('node:crypto');
const { spawnSync } = require('node:child_process');
const { cleanup, createTempDir, runNpm, isolatedNpmEnv } = require('./helpers.cjs');
const { SMOKE, runSmoke, CHILD_TIMEOUT_MS } = require('../scripts/release-tarball-smoke.cjs');
const { ensureHooksDist } = require('./helpers/hooks-dist.cjs');
const { SMOKE, runSmoke, entrypointFixtureHome, CHILD_TIMEOUT_MS, configuredEntrypointsIn } = require('../scripts/release-tarball-smoke.cjs');
const smokeMsg = (label, result) =>
`${label}: code=${result.code} details=${JSON.stringify(result.details)}`;
@@ -85,6 +87,11 @@ describe('release-tarball-smoke', () => {
let fixtureDir;
before(async () => {
// hooks/dist is gitignored and only produced by `npm run build:hooks`; the
// pack below must ship it or Cycle 4's entrypoint scan runs against a
// tarball with no hook scripts at all (SMOKE.INIT_FAILED on a clean tree).
ensureHooksDist();
// Pack once into a temp dir.
packDir = createTempDir('gsd-smoke-pack-');
installPrefix = createTempDir('gsd-smoke-prefix-');
@@ -436,6 +443,124 @@ describe('release-tarball-smoke', () => {
assert.ok(counts.after > 0 && counts.after < counts.before);
}
});
// ── H: configured entrypoints resolve for supported runtime profiles ──────
//
// #4154's scope bullet: the installer's in-process assertConfiguredEntrypoints
// gate is unit-covered in tests/configured-entrypoint-validation.test.cjs.
// This asserts the same property survives the packaged path — install the
// real tarball, run its installer for a runtime, and re-read that runtime's
// own config back off disk.
test('H: packed install leaves every configured entrypoint resolvable', () => {
const result = runSmoke({
tarballPath,
installPrefix,
expectedVersion: pkg.version,
fixtureDir,
// The lifecycle-command and workflow-body cycles are covered by A/C/E;
// skipping them here keeps this test to the entrypoint cycle.
lifecycleCommands: [],
entrypointRuntimes: ['claude', 'codex'],
npmEnv: isolatedNpmEnv(),
});
assert.equal(result.code, SMOKE.OK, smokeMsg('H', result));
assert.deepEqual(
result.details.entrypointProfiles.map((profile) => profile.runtime),
['claude', 'codex'],
smokeMsg('H', result),
);
for (const profile of result.details.entrypointProfiles) {
// Structural: a profile whose scan found nothing would satisfy the
// "no unresolved entrypoints" verdict vacuously.
assert.ok(
profile.entrypointsChecked > 0,
`${profile.runtime} configured no entrypoints: ${smokeMsg('H', result)}`,
);
}
});
// ── I: an unresolvable configured entrypoint fails the packed install ─────
//
// Red proof for ENTRYPOINT_UNRESOLVED, end to end through the packed tarball,
// and the reason this smoke check is not redundant with the installer's own
// gate: the fixture pre-registers a hook no install will ever write, the
// packed installer merges its own entries around it and exits 0 — its
// in-process assertConfiguredEntrypoints only validates paths the config
// writers registered with it during that run, so a registration it never
// touched is invisible to it — and reading settings.json back off disk is
// what catches the dangling launch path.
test('I: a registered hook that no install writes fails the smoke', (t) => {
const runtimeHome = entrypointFixtureHome(fixtureDir, 'claude');
// fixtureDir is shared with A/C/E/H, which install into this same home.
t.after(() => cleanup(runtimeHome));
const configDir = path.join(runtimeHome, '.claude');
const ghostHook = path.join(configDir, 'hooks', 'gsd-ghost-hook.js');
fs.mkdirSync(configDir, { recursive: true });
fs.writeFileSync(path.join(configDir, 'settings.json'), JSON.stringify({
hooks: {
PreToolUse: [
{ matcher: 'Bash', hooks: [{ type: 'command', command: `node "${ghostHook}"` }] },
],
},
}, null, 2));
const result = runSmoke({
tarballPath,
installPrefix,
expectedVersion: pkg.version,
fixtureDir,
lifecycleCommands: [],
entrypointRuntimes: ['claude'],
npmEnv: isolatedNpmEnv(),
});
assert.equal(result.code, SMOKE.ENTRYPOINT_UNRESOLVED, smokeMsg('I', result));
assert.equal(result.details.runtime, 'claude', smokeMsg('I', result));
assert.deepEqual(
result.details.unresolved,
[{ configPath: path.join(configDir, 'settings.json'), scriptPath: ghostHook }],
smokeMsg('I', result),
);
});
// ── J: configuredEntrypointsIn tolerates whitespace in configDir ──────────
//
// #4249 (antigravity review): the prior SCRIPT_PATH_RE excluded `\s` from
// the path match to avoid swallowing a shell command's trailing args, which
// also truncated any legitimate path containing a space — e.g. a real
// `/Users/John Doe/.claude` home — so the scan silently returned zero
// checked paths there. A direct unit test on the exported pure function:
// no packed install needed to prove this property.
test('J: configuredEntrypointsIn resolves a script path even when configDir contains a space', () => {
const configDir = path.join(os.tmpdir(), 'John Doe', '.claude');
const scriptPath = path.join(configDir, 'hooks', 'gsd-write-guard.js');
const text = JSON.stringify({
hooks: {
PreToolUse: [
{ matcher: 'Bash', hooks: [{ type: 'command', command: `node "${scriptPath}"` }] },
],
},
});
assert.deepEqual(configuredEntrypointsIn(text, configDir), [path.resolve(scriptPath)]);
});
// ── K: configuredEntrypointsIn does not swallow a preceding interpreter path ──
//
// #4249 (antigravity review): anchoring on the literal configDir prefix (a
// fix for J) must not regress the original "don't swallow the rest of a
// shell command" property — a command string that concatenates an
// interpreter path ahead of the real script path must resolve to the
// script path alone, not a combined interpreter+script string.
test('K: configuredEntrypointsIn does not include a preceding interpreter path', () => {
const configDir = path.join(os.tmpdir(), '.claude');
const scriptPath = path.join(configDir, 'hooks', 'gsd-write-guard.js');
const text = `"command": "/usr/local/bin/node ${scriptPath} --flag"`;
assert.deepEqual(configuredEntrypointsIn(text, configDir), [path.resolve(scriptPath)]);
});
});

View File

@@ -3135,8 +3135,14 @@ describe('partitionIsolatedFiles (#4497 codex-config.test.cjs chunk isolation, d
// same set as each other under the OLD (pre-fix) platform-scaled rule this
// row is guarding against — confirming both rows are load-bearing, not
// vacuous.
//
// #4249 split codex-config.test.cjs's heavy install()-pipeline blocks into
// codex-config-hooks.test.cjs, dropping codex-config.test.cjs's own
// measured weight from 127783ms to 189ms — well under this derived bar
// (0.3 * CHUNK_WORKING_BUDGET_MS ~= 120000ms). It correctly no longer
// appears in the live-computed isolated set, so it is dropped from this
// pinned expectation too, in place of being isolated by name.
const EXPECTED_ISOLATED_UNIT_FILES = [
'codex-config.test.cjs',
'config.test.cjs',
'emitted-attribution.test.cjs',
'install-minimal-hooks.test.cjs',

View File

@@ -4,9 +4,10 @@
"unit": "ms",
"sources": [
"test-events-linux-node22.jsonl",
"test-events-linux-node24.jsonl"
"test-events-linux-node24.jsonl",
"manual: codex-config.test.cjs/codex-config-hooks.test.cjs re-measured locally (node --test --test-reporter=tap, max of 3 runs) after next split codex-config.test.cjs (#4139/#4540) — the old 127783ms entry predated the split and codex-config-hooks.test.cjs, which now holds the #3245/#4249 install()-pipeline blocks, had no entry at all (PR #4249)."
],
"file_count": 770,
"file_count": 771,
"timings": {
"active-workstream-store.test.cjs": 81,
"active-workstream-store.unit.test.cjs": 364,
@@ -129,7 +130,8 @@
"codebase-mapper-date-restamp.test.cjs": 138,
"codebuddy-install.test.cjs": 3763,
"codebuddy-upgrades.test.cjs": 2932,
"codex-config.test.cjs": 127783,
"codex-config-hooks.test.cjs": 3454,
"codex-config.test.cjs": 189,
"codex-declarative-reference.test.cjs": 205,
"command-arg-projection.test.cjs": 113,
"command-contract.test.cjs": 1105,