fix(#370): hook wrappers silent no-op outside an apexyard fork by atlas-apex · Pull Request #371 · me2resh/apexyard

atlas-apex · 2026-05-21T16:45:19Z

Summary

Fixes the UI-flooding "No such file or directory" error that every .claude/settings.json hook wrapper produced on every tool call when Claude Code was launched outside an apexyard ancestry (e.g. /tmp, a bare project clone, any unrelated repo). Root cause: the walk-up loop correctly stops at / when no anchor file is found, but the wrapper unconditionally execs $r/.claude/hooks/<name>.sh — which lands on /.claude/hooks/<name>.sh (doesn't exist). One error per wrapper, per tool call → unusable terminal.
Inserts an anchor-found guard between done; and exec in all 43 wrappers: done;[ -f "$r/.apexyard-fork" ] || [ -f "$r/onboarding.yaml" ] || exit 0;exec .... When neither anchor is found, the wrapper silently exits 0 — the correct behaviour outside an ops fork (hooks are framework-internal and shouldn't fire on unrelated repos).
Single uniform transform applied via a python JSON-safe text substitution: 43 OLD → 43 NEW, JSON re-validated post-edit. Zero unguarded done;exec patterns remain in .claude/settings.json.
New regression test test_settings_wrappers_silent_noop.sh locks in three invariants: every wrapper has the guard (catches future-added wrapper without the guard), wrapper invoked from /tmp exits 0 silently (empirical bug-fix confirmation), wrapper invoked from inside the fork on a non-trigger path still exits 0 (negative control proving the guard doesn't accidentally block legitimate invocations).

Testing

bash .claude/hooks/tests/test_settings_wrappers_silent_noop.sh — PASS at 3/3.
Manual reproduction: (cd /tmp && bash -c '<wrapper>') — exits 0 with empty stderr (was: bash: /.claude/hooks/<name>.sh: No such file or directory).
In-fork invocation still works: PreToolUse Edit hook on a non-security-sensitive path → exits 0 silently as before.
All existing hooks tests still PASS (no regression in the rest of the test suite — the fix is in the wrapper, not in any hook's body).

Glossary

Term	Definition
Hook wrapper	The `bash -c '...'` invocation in `.claude/settings.json` that wraps each `.claude/hooks/<name>.sh` call. Pre-#370 shape: walk up from `$PWD` to find an ops-fork anchor, then exec the hook at the resolved path. Post-#370 shape: same walk-up, but if no anchor is found, silently exit 0 instead of attempting to exec at `/`.
Anchor file	One of `.apexyard-fork` (v2 marker) or `onboarding.yaml` (v1 legacy). The wrapper resolves the ops-fork root by walking up the directory tree looking for either. Per AgDR-0021 § B + AgDR-0041 § "Decision" point 2 — wrappers accept either form.
Silent no-op outside the fork	The corrected behaviour. Adopter runs Claude Code in `/tmp` or a bare clone, the framework's hooks don't fire (correct — they're framework-internal), and the operator's terminal stays clean. Previously: noise flood per tool call.
Walk-up loop termination at `/`	Bash idiom: `while ... && [ "$r" != / ]; do r=${r%/*}; done`. When no anchor is found, `$r` arrives at `/`. The fix's `

Closes #370

When Claude Code is launched from a directory with no apexyard ancestry (e.g. /tmp, a bare project clone, or any dir outside the ops fork), all 43 hook wrappers crashed with "No such file or directory" on every tool call: Failed with non-blocking status code: bash: /.claude/hooks/detect-role-trigger.sh: No such file or directory Root cause: the walk-up loop in each wrapper correctly stops at / when no anchor file is found, but the wrapper unconditionally execs $r/.claude/hooks/<name>.sh — landing on /.claude/hooks/<name>.sh when $r is /. One error per wrapper, per tool call → UI noise flood. Fix: insert an anchor-found guard between `done;` and `exec`: done;[ -f "$r/.apexyard-fork" ] || [ -f "$r/onboarding.yaml" ] || exit 0;exec ... When neither anchor is found, the wrapper exits 0 silently — the correct behaviour outside an ops fork (hooks are framework-internal and shouldn't fire on unrelated repos). Scope: all 43 wrappers in .claude/settings.json. Applied via a python JSON-safe text substitution + re-validated the JSON parses cleanly. Regression test (test_settings_wrappers_silent_noop.sh): - Invariant 1: every wrapper has the '|| exit 0;exec' guard (no unguarded `done;exec` patterns) — protects against a regression where a new wrapper added later forgets the guard - Invariant 2: a representative wrapper invoked from /tmp (no fork ancestry) exits 0 with empty stderr — empirical confirmation of the bug fix - Invariant 3: same wrapper invoked from inside the fork on a non- trigger path still exits 0 silently — negative control proving the guard doesn't accidentally block legitimate in-fork invocations 3/3 PASS at HEAD. Closes #370 Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

atlas-apex

Code Review: PR #371

Commit: 4b426ab780aca4aa9f4350df477edd0d7155b5c5

Verdict intent: CHANGES REQUESTED. Filed as a comment review because GitHub disallows self---request-changes. The blocking item below is non-advisory and must be resolved before merge.

Summary

Targeted bug fix for the UI-flooding "No such file or directory" noise that every .claude/settings.json hook wrapper produced on every tool call when Claude Code was launched outside an apexyard ancestry. Root cause is correctly diagnosed: walk-up loop terminates at / when no anchor is found, but the wrapper unconditionally execs $r/.claude/hooks/<name>.sh, landing on /.claude/hooks/<name>.sh (doesn't exist). Fix inserts an anchor-found guard ([ -f "$r/.apexyard-fork" ] || [ -f "$r/onboarding.yaml" ] || exit 0) between done; and exec in all 43 wrappers. Scope is exactly 2 files (settings.json + new regression test). The fix is correct, the bug reproduces pre-fix and is silenced post-fix.

Checklist Results

Architecture & Design: Pass (single-anchor laxity at wrapper level matches AgDR-0041 § Decision point 2; both-files-required strictness stays in the hook-internal resolver)
Code Quality: Pass (uniform transform, JSON re-validated)
Testing: Pass with one strengthening suggestion (see below)
Security: N/A (no auth/crypto/secrets touched)
Performance: Pass (one extra -f test per wrapper invocation, negligible)
PR Description & Glossary: Pass (narrative summary bullets, 4-row glossary)
Technical Decisions (AgDR):N/A (faithful execution of existing AgDR-0021 + AgDR-0041; no new decision)
Adopter Handbooks: N/A (no architecture / migration / language-specific findings)

Issues Found

⛔ BLOCKING: CI shellcheck workflow failing at HEAD 4b426ab — https://github.com/me2resh/apexyard/actions/runs/26239945147

The new regression test trips SC2164 twice:

./.claude/hooks/tests/test_settings_wrappers_silent_noop.sh:71:3: warning: Use 'cd ... || exit' or 'cd ... || return' in case cd fails. [SC2164]
./.claude/hooks/tests/test_settings_wrappers_silent_noop.sh:89:3: warning: Use 'cd ... || exit' or 'cd ... || return' in case cd fails. [SC2164]

The shellcheck .claude/hooks workflow runs at -S warning, so warnings block. Per .claude/rules/pr-quality.md § "No Red CI Before Merge" — never merge with red CI, even pre-existing or unrelated. Here the failure is new and caused by this PR.

Fix:

# line 71
cd /tmp || exit 1

# line 89
cd "$ROOT" || exit 1

Verification (positive findings)

Uniform transform correctness:
- grep -cF '|| exit 0;exec' .claude/settings.json → 43 ✓
- grep -cF 'done;exec' .claude/settings.json → 0 ✓
- grep -cF 'r=$PWD' .claude/settings.json → 43 ✓
- Unique JSON-escaped transform shape: 1 ✓
- No drive-by edits — every +/- in settings.json is wrapper-shape change
JSON validity post-edit: jq . .claude/settings.json > /dev/null exits 0 ✓
Guard semantics: walk-up loop and guard use the same || shape (single-anchor laxity at the wrapper level, both-files-required stays in the in-hook resolver). Matches AgDR-0041 § Decision point 2. Sanity test: injected a .apexyard-fork anchor in /tmp/XXX/, ran the wrapper from there → hook DID run (stderr HOOK_RAN, rc=0). Confirms the guard ALLOWS exec when an anchor exists.
Bug reproduction:
- Pre-fix shape (no guard) run from /tmp: rc=126, stderr = bash: /.claude/hooks/detect-role-trigger.sh: No such file or directory; bash: line 0: exec: ... cannot execute: No such file or directory
- Post-fix shape (this PR) run from /tmp: rc=0, empty stderr
- Fix works as advertised.
Regression test 3/3 PASS locally.
Scope discipline: 2 files, +154/-43. No other edits.

Suggestions

(nit) Strengthen Invariant 3 with a positive in-fork case. Current invariant 3 runs the wrapper inside the fork on a non-trigger path and asserts rc=0 + empty stderr. This passes whether the hook actually ran (silent on non-trigger) OR if the guard accidentally exits 0 always (hook never runs). The 3-invariant suite as a whole is fine — invariant 1 enforces the guard SHAPE textually, which makes the "always exits" failure mode impossible to reach without also tripping invariant 1. Still, an additional positive sub-case ("inside fork on a TRIGGER path → expected banner appears on stderr") would prove end-to-end behaviour rather than relying on the textual shape check. Worth a follow-up, not blocking.
(nit) Test scope: one wrapper as proxy for 42. The bug is in the wrapper SHAPE, not in any individual hook body, and invariant 1 (textual count check) catches the shape regression for all 43. The runtime smoke tests only detect-role-trigger.sh. Acceptable as-is given the uniform transform; a per-wrapper smoke would yield diminishing returns. PR-acceptable.
(suggestion) Wire test_settings_wrappers_silent_noop.sh into CI. The framework's convention is per-test CI wiring (see extract-subpacks-on-release.yml, site-counts-check.yml). Without a CI entry, invariant 1 will only be checked when someone manually runs the test. A small dedicated workflow (or a generic run-hook-tests.yml that globs .claude/hooks/tests/test_*.sh) would close the loop. Not blocking; can ship as a follow-up.

Verdict

CHANGES REQUESTED (filed as comment per self-PR constraint)

The fix itself is correct, well-scoped, and well-tested locally. The blocker is the new red CI status from the SC2164 shellcheck warnings on the test file — fixable in ~30 seconds (|| exit 1 on the two cd calls). Once CI is green at the updated HEAD, this is approval-ready and I'd recommend re-running /code-review on the new commit.

🤖 Reviewed by Rex (Code Reviewer Agent)
📌 Reviewed commit: 4b426ab780aca4aa9f4350df477edd0d7155b5c5

Rex review on PR #371 caught two SC2164 warnings on the new test that the framework's shellcheck CI workflow blocks at `-S warning`. Both calls were inside subshells (so `cd` failure wouldn't propagate to the test loop) — but shellcheck doesn't analyse subshell scope. Add explicit `|| exit 1` guards to satisfy the linter. Line 71: `cd /tmp || exit 1` Line 89: `cd "$ROOT" || exit 1` Test still 3/3 PASS locally. Refs #370 PR #371 Rex review (CI red blocker). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

atlas-apex

Code Review: PR #371 (re-review after shellcheck fix)

Commit: f486e4ade431251e473483f7ade8a34a41d1a3b0

Summary

Re-review at new HEAD f486e4a after the shellcheck SC2164 follow-up. Prior review (CHANGES REQUESTED on 4b426ab for CI red on SC2164 warnings at test_settings_wrappers_silent_noop.sh:71 + :89) is now resolved.

Verification against the 5 checks

#	Check	Status
1	`cd ... \|\| exit 1` is the canonical SC2164 fix	PASS — `
2	Test still 3/3 PASS	PASS — ran `bash .claude/hooks/tests/test_settings_wrappers_silent_noop.sh` at HEAD `f486e4a`: 3 passed, 0 failed
3	`shellcheck .claude/hooks` CI workflow green at HEAD	PASS — workflow completed SUCCESS at 2026-05-21T17:03:22Z (was the red blocker on prior HEAD)
4	No drive-by edits — diff is exactly the 2 lines	PASS — `git diff 4b426ab..f486e4a --stat` reports `1 file changed, 2 insertions(+), 2 deletions(-)` on `test_settings_wrappers_silent_noop.sh`; lines 71 + 89 only
5	Original PR main work unchanged	PASS — at HEAD: 43 wrappers (`r=$PWD` count), 43 guarded (`\|\| exit 0;exec` count), 0 unguarded (`done;exec` count). Matches the prior approved-modulo-CI state.

Checklist Results

Architecture & Design: Pass (no architectural surface change; pure CI hygiene)
Code Quality: Pass
Testing: Pass — regression test still asserts the three invariants and now lints clean
Security: Pass (no security surface)
Performance: Pass (no performance surface)
PR Description & Glossary: Pass (inherited from the prior commit; follow-up commit message clearly explains the why)
Summary Bullet Narrative: Pass
Technical Decisions (AgDR):N/A — pure linter-hygiene fix, no decision surface
Adopter Handbooks: N/A — no handbooks loaded for shell-test diffs

Issues Found

None.

Suggestions

None.

Verdict

APPROVED — all three CI checks green, test passes, scope is exactly the 2 lines described, original 43-wrapper work intact. Cleared for merge from the code-review side; operator owns the marker write + per-PR CEO approval.

Reviewed by Rex (Code Reviewer Agent)
Reviewed commit: f486e4ade431251e473483f7ade8a34a41d1a3b0

* fix(#370): hook wrappers silent no-op outside an apexyard fork When Claude Code is launched from a directory with no apexyard ancestry (e.g. /tmp, a bare project clone, or any dir outside the ops fork), all 43 hook wrappers crashed with "No such file or directory" on every tool call: Failed with non-blocking status code: bash: /.claude/hooks/detect-role-trigger.sh: No such file or directory Root cause: the walk-up loop in each wrapper correctly stops at / when no anchor file is found, but the wrapper unconditionally execs $r/.claude/hooks/<name>.sh — landing on /.claude/hooks/<name>.sh when $r is /. One error per wrapper, per tool call → UI noise flood. Fix: insert an anchor-found guard between `done;` and `exec`: done;[ -f "$r/.apexyard-fork" ] || [ -f "$r/onboarding.yaml" ] || exit 0;exec ... When neither anchor is found, the wrapper exits 0 silently — the correct behaviour outside an ops fork (hooks are framework-internal and shouldn't fire on unrelated repos). Scope: all 43 wrappers in .claude/settings.json. Applied via a python JSON-safe text substitution + re-validated the JSON parses cleanly. Regression test (test_settings_wrappers_silent_noop.sh): - Invariant 1: every wrapper has the '|| exit 0;exec' guard (no unguarded `done;exec` patterns) — protects against a regression where a new wrapper added later forgets the guard - Invariant 2: a representative wrapper invoked from /tmp (no fork ancestry) exits 0 with empty stderr — empirical confirmation of the bug fix - Invariant 3: same wrapper invoked from inside the fork on a non- trigger path still exits 0 silently — negative control proving the guard doesn't accidentally block legitimate in-fork invocations 3/3 PASS at HEAD. Closes #370 Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(#370): shellcheck SC2164 — guard cd calls in regression test Rex review on PR #371 caught two SC2164 warnings on the new test that the framework's shellcheck CI workflow blocks at `-S warning`. Both calls were inside subshells (so `cd` failure wouldn't propagate to the test loop) — but shellcheck doesn't analyse subshell scope. Add explicit `|| exit 1` guards to satisfy the linter. Line 71: `cd /tmp || exit 1` Line 89: `cd "$ROOT" || exit 1` Test still 3/3 PASS locally. Refs #370 PR #371 Rex review (CI red blocker). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> --------- Co-authored-by: me2resh <ahmed.abdelaliem@gmail.com> Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

github-actions Bot added github-issue has-ticket labels May 21, 2026

atlas-apex commented May 21, 2026

View reviewed changes

atlas-apex merged commit 181c1a4 into dev May 21, 2026
3 checks passed

atlas-apex deleted the fix/GH-370-hook-wrappers-silent-noop-outside-fork branch May 21, 2026 17:22

atlas-apex mentioned this pull request May 21, 2026

fix: hook wrappers crash with 'No such file or directory' when launched outside an apexyard fork #370

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix(#370): hook wrappers silent no-op outside an apexyard fork#371

fix(#370): hook wrappers silent no-op outside an apexyard fork#371
atlas-apex merged 2 commits into
devfrom
fix/GH-370-hook-wrappers-silent-noop-outside-fork

atlas-apex commented May 21, 2026

Uh oh!

atlas-apex left a comment

Uh oh!

atlas-apex left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

atlas-apex commented May 21, 2026

Summary

Testing

Glossary

Uh oh!

atlas-apex left a comment

Choose a reason for hiding this comment

Code Review: PR #371

Summary

Checklist Results

Issues Found

Verification (positive findings)

Suggestions

Verdict

Uh oh!

atlas-apex left a comment

Choose a reason for hiding this comment

Code Review: PR #371 (re-review after shellcheck fix)

Summary

Verification against the 5 checks

Checklist Results

Issues Found

Suggestions

Verdict

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants