The windows-test-parity ratchet greps test source for fs.rmSync-without-
maxRetries (and six other Windows-portability anti-patterns), failing when an
integer offender COUNT exceeds a frozen baseline (rmSync: 95). A count ratchet
is a Goodhart metric: fixing one offender and adding another keeps the count
constant, so a new defect slips through green. Replace it — and every other
count ratchet in the repo — with a layered, masking-proof design.
Behavioral seam test
- tests/helpers-cleanup.test.cjs proves helpers.cleanup() carries the Windows
EBUSY retry budget. cleanup() delegates retries to Node's fs.rmSync via
maxRetries (it owns no loop), so the test asserts the option contract
(recursive/force/maxRetries>0/retryDelay>0) + real-FS removal + the cwd-guard,
rather than a loop that does not exist. The EBUSY risk is now tested ONCE at
the helper, not approximated textually at every call site.
Write-time ESLint rule (AST-accurate, replaces the grep)
- eslint-rules/no-raw-rmsync-in-tests.cjs (error in tests/**/*.test.cjs) bans
raw fs.rmSync, steering to cleanup(). Catches member, computed (fs['rmSync']),
destructured and aliased forms; escape hatch is inline
`// eslint-disable-next-line local/no-raw-rmsync-in-tests -- <reason>` only.
- Migrated 336 raw fs.rmSync teardown calls across ~116 test files to cleanup().
~18 genuinely load-bearing sites (mid-test SUT/fault-injection removals,
error-swallowing or name-colliding local teardown helpers) keep the raw call
with an inline eslint-disable + reason.
Shared anti-ratchet primitive
- scripts/lib/allowlist-ratchet.cjs:
- assertWithinAllowlist: fails on NOVEL ids (new offender introduced) AND on
STALE ids (a known offender was fixed but not pruned) — identity, not count,
and a ratchet DOWN toward zero.
- assertTightCeiling: a size/length budget whose ceiling must stay within a
grace band of the high-water mark, so budgets may only tighten, never creep.
Ratchets converted onto the primitive
- windows-test-parity-guard.test.cjs: rmSync rule deleted (now ESLint-enforced);
the remaining six patterns moved from integer baselines to named-set
allowlists with ratchet-down.
- scripts/lint-test-file-count.{cjs,allowlist.json}: per-module integer counts →
named filename sets (closes the swap-a-file-keep-the-count blind spot); a
module dropping under cap now FAILS to force pruning its allowlist entry.
- enh-2790 skill-count `<= 63` → named skill allowlist (ratchets toward ~58).
Size budgets hardened (tighten-only)
- agent-size / workflow-size / feat-3039 help-tiered: ceilings lowered to the
current high-water mark and an assertTightCeiling anti-creep check added per
tier. Fixed external-contract limits (description ≤100 chars, agent ≤100 KB)
are intentionally left as-is — they are not grandfathered creeping budgets.
No user-facing behavior change (tests + tooling only); no USER_FACING_PREFIXES
touched, so no changeset fragment is required.
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* test(3659): add regression tests for applySurface skill-dir pruning on cluster disable
Tests that applySurface with claude global scope correctly prunes
~/.claude/skills/gsd-STEM/ dirs for disabled clusters, preserves
gsd-STEM dirs in enabled clusters, leaves non-gsd user dirs untouched,
and is idempotent across two consecutive calls.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(3659): applySurface now prunes ~/.claude/skills/gsd-STEM/ on cluster disable
Root cause: surface.md directed the AI to use RUNTIME_CONFIG_DIR=~/.claude/skills
(the skills sub-directory) instead of the base Claude config dir (~/.claude).
When runtimeConfigDir=~/.claude/skills and scope=global, the layout computes
dest=~/.claude/skills/skills — the wrong target — so pruning never reached
the actual gsd-STEM dirs in ~/.claude/skills/.
Fix:
- surface.md: correct RUNTIME_CONFIG_DIR to use the base config dir (~/.claude),
add explicit SCOPE=global, and update all path references in execution_context.
Surface state file moves from ~/.claude/skills/.gsd-surface.json to
~/.claude/.gsd-surface.json, matching install/uninstall conventions.
- surface.cjs: extract pruneSkillDirs() as a shared helper (single point of truth
for gsd-STEM dir removal). _syncGsdDir now delegates to it instead of having
the ownership/prune logic inline. Export pruneSkillDirs for callers that need
stand-alone pruning without a full applySurface pass.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* chore(3659): update changeset to reference PR #3766
* fix(3659): manifest-membership gate on pruneSkillDirs prevents user gsd-* dir data loss
Finding 1 (CRITICAL): the prefixed branch previously deleted any on-disk dir that
matched the 'gsd-' prefix and was not in retainedNames. A user-created gsd-mything/
would be silently destroyed. Fix: deletion now requires BOTH prefix match AND
manifest membership (stem present in manifest). Dirs that match the prefix but are
not manifest-known are preserved with a process.stderr warning so the user knows the
dir was kept.
Finding 2 (type guard): the Hermes (empty-prefix) branch passed manifest directly to
new Set([...manifest.keys()]) without verifying it is actually a Map. A truthy
non-Map would throw. Fix: safeManifest = (manifest instanceof Map) ? manifest : null,
used in both branches. Non-Map manifest triggers the same conservative
no-deletions path already used when manifest is absent.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* test(3659): counter-test for all-clusters-disabled + user gsd-* dir preservation
Finding 3: add test (e) that disables every cluster (Object.keys(CLUSTERS)) and
asserts three things:
1. All GSD-owned skill dirs (gsd-explore/, gsd-help/) are removed.
2. Non-gsd user dir (my-custom-skill/) is preserved.
3. User-created gsd-mything/ (prefix match, not in manifest) is preserved —
this is the critical regression guard for the Finding 1 data-loss fix.
Also update the existing _syncGsdDir skills-kind test in surface-apply.test.cjs to
pass a manifest that declares old-skill as GSD-owned. Without a manifest the new
conservative path correctly preserves all unknown gsd-* dirs, which broke the
pre-existing no-manifest assertion; supplying the manifest restores the expected
pruning behavior and documents the required calling contract.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(3659): address pr-review-toolkit + codex review findings
- Collapse redundant if/else in _syncGsdDir
- Collapse duplicate canonicalStems branches in pruneSkillDirs
- Update stale module-header comment (config-dir root)
- Clarify dead isGsdOwned guard comment
- Log rmSync failures to stderr
- Add pruneSkillDirs to module-header Exports JSDoc
- Remove unused imports in bug-3659 test file
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
