fix(source-control): actionable owner/repo diagnostic on readiness-gate fetch failure by kyle-sexton · Pull Request #839 · melodic-software/claude-code-plugins

kyle-sexton · 2026-07-21T09:35:00Z

Summary

babysit-readiness-gate.sh <N> (safe-tier babysit) could exit 4 with only a
generic fetch-all-pr-comments.sh failed for PR <N> line — the same exit code
as a missing jq — leaving a worker following the documented invocation to
rediscover, by trial and error, that setting FETCH_COMMENTS_OWNER /
FETCH_COMMENTS_REPO fixes it. This makes that failure mode self-explanatory
and documents the override in the worker contract.

Fix

Root cause: the gate shells out to fetch-all-pr-comments.sh, which
auto-derives owner/repo from the current directory via gh repo view. The
auto-derive already works from a checkout of the target repo; it returns empty
only when the cwd is not such a checkout (e.g. a targeted-recheck pass that
never runs gh pr checkout). That is inherent to cwd-based resolution — the
FETCH_COMMENTS_OWNER/FETCH_COMMENTS_REPO override is the cwd-independent
escape hatch, and it already worked; it was just undocumented and the failure
was undifferentiated. So this change is diagnostics + docs, with no behavior
change to the readiness verdict or exit codes:

Gate diagnostic — on fetch failure the gate now prints the resolved cwd
and names the FETCH_COMMENTS_OWNER/FETCH_COMMENTS_REPO override (and the
"run from the target repo checkout" fix), in addition to fetch-all-pr-comments.sh's
own stderr. Its --help gains a "Repo resolution" section and the exit-4
doc now covers the fetch-failure case, not just missing jq.
Fetch diagnostic — fetch-all-pr-comments.sh's cannot resolve owner/repo
message now names the cwd gh repo view was resolved from and the override.
Worker contract — SKILL.md (Step 0.1) and reference/loop.md (§5.0.3
fetch step, §5.1.3 gate step E) document the override as a prerequisite for
passes whose cwd is not the target-repo checkout.

Considered and rejected: a new --repo owner/repo flag. The env-var override
already provides cwd-independent resolution, so a flag would duplicate a working
mechanism for no gain; keeping the surface small.

Verification

Reproduced the exit 4 from a non-repo cwd, then confirmed the new
diagnostic names the cwd and both env vars, still exiting 4.
babysit-readiness-gate.test.sh — added a gate-level case: with a
gh repo view→empty stub on PATH and no --comments-json, the gate exits
4 and its stderr names FETCH_COMMENTS_OWNER; plus a --help-documents-the-override
assertion. 63/63 pass.
fetch-all-pr-comments.test.sh — added: unresolvable owner/repo exits 2
with the actionable message, and the env-var override resolves owner/repo even
when gh repo view is empty. 17/17 pass.
shellcheck clean on all four scripts; skill-quality check babysit-prs
PASS (SKILL.md 499/500, markdownlint clean).

Closes source-control:babysit-prs: babysit-readiness-gate.sh exits 4 without FETCH_COMMENTS_OWNER/FETCH_COMMENTS_REPO env vars, undocumented in worker contract #475.
source-control:babysit-prs: babysit-readiness-gate.sh over-counts lifetime P-badges as 'findings', false READINESS_BLOCKED reason=under-decomposed on fully-classified PRs #465 — readiness gate over-counts lifetime P-badges as findings (a separate
gate counting bug; context only, not fixed here).
source-control:babysit-prs: read-only babysit_merge.py invocation blocked by host auto-mode permission classifier #787 — babysit_merge.py read-only invocation blocked by the host auto-mode
permission classifier (sibling babysit-prs issue; context only, not fixed here).

Closes #475

🤖 Generated with Claude Code

https://claude.ai/code/session_01K1V3gkrfSf75isB8MiDy3o

…ailure babysit-readiness-gate.sh shells out to fetch-all-pr-comments.sh, which auto-derives owner/repo from the current directory via `gh repo view`. From a cwd that is not a checkout of the target repo (e.g. a targeted-recheck pass) that derivation returns empty and the fetch exits non-zero, which the gate surfaced only as a generic "fetch-all-pr-comments.sh failed" line plus exit 4 — the same undocumented exit code as a missing jq, leaving a worker to rediscover the FETCH_COMMENTS_OWNER/FETCH_COMMENTS_REPO override by trial and error. The gate's failure message now names the resolved cwd and the env-var override, fetch-all-pr-comments.sh's own "cannot resolve owner/repo" message names the cwd and override, and the gate's --help and exit-code docs cover it. The babysit-prs worker contract (SKILL.md, reference/loop.md) documents the override as a prerequisite for out-of-checkout passes. Regression tests cover the gate's exit-4 diagnostic, the fetch script's exit-2 message, and the env-var override path. No change to the readiness verdict or exit codes; diagnostics and docs only. Closes #475 Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01K1V3gkrfSf75isB8MiDy3o

claude · 2026-07-21T09:35:14Z

Claude finished @kyle-sexton's task in 2m 51s —— View job

Code Review

Gather context and understand the PR
Review changed files for correctness, security, and guidelines alignment
Post review findings

Verdict: Approve. This is a well-scoped, diagnostics-only fix with solid test coverage and no behavior change to the readiness verdict or exit codes. A few nits below, none blocking.

`babysit-readiness-gate.sh`

usage() line range is correctly updated from sed -n '2,58p' → sed -n '2,65p'. The delta of 7 accounts for the 6-line "Repo resolution" section plus the 1 extra exit-code doc line, and line 65 (# for the cause...) is the last comment line before set -uo pipefail on line 67. ✓

Diagnostic message safety — both new printf calls use %s with "$PWD" as a positional argument rather than interpolating $PWD into the format string. No word-splitting or glob risk.

# ✓ safe
printf '...current directory (%s)...\n' "$PWD" >&2

Env var passthrough — FETCH_COMMENTS_OWNER/FETCH_COMMENTS_REPO are inherited by the child bash "$SCRIPT_DIR/fetch-all-pr-comments.sh" process without needing explicit -v passing; export in the caller is sufficient and matches what the docs instruct. ✓

`fetch-all-pr-comments.sh`

One-line change extending the "cannot resolve owner/repo" message with $PWD and the override names. Same safe printf/%s pattern. ✓

`babysit-readiness-gate.test.sh`

New stub + assertions look correct.

$* matching in the stub — case "$*" in *"repo view"*) concatenates all args. For gh repo view --json owner -q .owner.login the string includes repo view and the pattern fires correctly. ✓

Exit-code capture pattern:

fetch_rc=$(PATH="$NOREPO_BIN:$PATH" bash "$GATE" 123 >/dev/null 2>&1; echo $?)

The >/dev/null 2>&1 suppresses output from the gate; echo $? then writes the exit code to the subshell stdout, which $(...) captures. Correct — $? here is the gate's exit, not echo's. ✓

--help assertion — checks that FETCH_COMMENTS_OWNER appears in help output, which ensures the new "Repo resolution" section is exposed via --help. ✓

`fetch-all-pr-comments.test.sh`

Case 9 and 10 share the same stub ($TEST_TMPDIR/bin-norepo/gh) created in Case 9's setup block. Fine in a sequential shell script but worth noting as an implicit dependency if tests are ever reordered.

Case 10 only asserts exit 0, not output content. Since the stub returns [] for non-repo view gh calls, the fetch script should produce a valid empty array — exit 0 is the meaningful signal for "override unblocked resolution." Light but sufficient for the stated goal. ✓

`SKILL.md` — Minor readability nit

Step 0.1 now embeds a long parenthetical inline:

fetch-all-pr-comments.sh (derives owner/repo from the current directory; from a cwd that is not a checkout of the target repo, export ... first — also unblocks the readiness gate's exit 4), filter own prior replies, classify

This makes Step 0.1 a run-on sentence. Markdownlint passes and the information is correct; just harder to scan. Could be a follow-up to break the override note into its own sub-bullet. Not blocking.

`reference/loop.md`

§5.0.3 and §5.1.3 additions are clear, accurate, and consistent with what the scripts now actually emit. The §5.0.3 note quotes the old "cannot resolve owner/repo" fragment (the new message is longer), but that's a reasonable summary rather than a literal quote. ✓

Other

plugin.json 0.15.7 → 0.15.8 is the correct patch bump for a diagnostics-only fix. ✓
CHANGELOG.md entry is thorough and accurate. ✓
shellcheck clean on all four scripts per the PR's verification. ✓

claude · 2026-07-21T09:35:20Z

Claude finished @kyle-sexton's task in 1m 42s —— View job

Security Review

Gather context — reading changed files
Analyze for security issues
Post findings

No security issues found.

This PR is diagnostics and documentation only — it adds two printf error messages, updates one existing printf message, adds test cases, and updates markdown docs. Here is what I checked:

$PWD in diagnostic printf calls (babysit-readiness-gate.sh L152, fetch-all-pr-comments.sh L101): $PWD is passed as a positional %s argument, not interpolated into the format string — no format-string injection surface. A crafted directory name containing ANSI control sequences could corrupt terminal output, but this is a theoretical-only risk limited to stderr diagnostics, carries no code-execution or privilege-escalation vector, and is consistent with patterns found throughout the codebase.

FETCH_COMMENTS_OWNER / FETCH_COMMENTS_REPO env vars documented as the override path: the var-reading code and its use inside gh api "repos/$OWNER/$REPO/..." calls predate this PR entirely; no new trust boundary is opened here.

Test stubs (both test files): each stub gh binary is written via a single-quoted heredoc into a $TEST_TMPDIR subdirectory and prepended to PATH only for the specific test invocation — standard safe pattern.

Other changes (plugin.json version bump, SKILL.md, reference/loop.md): documentation only, no executable surface.

kyle-sexton · 2026-07-21T09:39:48Z

Post-green review: Approve/clean per both automated passes. No blocking findings — one non-blocking readability nit (SKILL.md run-on sentence in Step 0.1) left alone per the reviewer's own "not blocking" call. No inline review threads. Handing off to babysit for ongoing watch.

chatgpt-codex-connector · 2026-07-21T09:39:52Z

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, you can upgrade your account or add credits to your account and enable them for code reviews in your settings.

claude · 2026-07-21T09:40:04Z

Claude finished @kyle-sexton's task in 2m 52s —— View job

Code Review

Gather context and understand the PR
Review changed files for correctness, security, and guidelines alignment
Post review findings

Verdict: Approve. Diagnostics-only fix with correct shell idioms, thorough test coverage, and no behavior change to the readiness verdict or exit codes. A few observations below — none blocking.

`babysit-readiness-gate.sh`

sed -n '2,65p' range — Confirmed correct. The diff adds a 6-line "Repo resolution" block (lines 50–55) and expands the exit-4 doc to two lines (64–65); line 65 is the last #-prefixed comment before set -uo pipefail on line 67. Delta of 7 from the old 2,58p matches. ✓

Diagnostic ordering — The gate runs bash "$SCRIPT_DIR/fetch-all-pr-comments.sh" and the child script's stderr appears (via inherited fd 2) before the || handler fires. The "(fetch-all-pr-comments error above)" phrasing is accurate in practice. ✓

$PWD in printf — Both new printf calls pass $PWD as a %s positional argument, not via format-string interpolation. No injection surface. ✓

# ✓ safe — $PWD is a positional arg, not in the format string
printf '...current directory (%s)...\n' "$PWD" >&2

`fetch-all-pr-comments.sh`

One-line extension of the "cannot resolve owner/repo" error at L101. Same safe printf/%s pattern, names cwd and both env vars. ✓

`babysit-readiness-gate.test.sh`

$* stub matching — case "$*" in *"repo view"*) concatenates all args; for gh repo view --json owner -q .owner.login the string includes repo view and the glob fires correctly. ✓

PATH inheritance — PATH="$NOREPO_BIN:$PATH" bash "$GATE" 123 exports the modified PATH into the gate process; when the gate in turn runs bash "$SCRIPT_DIR/fetch-all-pr-comments.sh", that child inherits the same PATH, so the stub gh intercepts repo view inside the fetch script too. ✓

Exit-code capture pattern — $(... bash "$GATE" 123 >/dev/null 2>&1; echo $?) correctly captures the gate's exit via echo $? after the suppressed command. ✓

Gate runs twice — exit code and stderr are captured in two separate subshells. This is fine for a test but means the gate (and its child gh api user call against the stub) runs twice. Minor inefficiency, not a correctness issue. If the test suite grows slow this is a candidate for a single run with a temp file.

--help assertion — assert_contains "--help documents the env-var override" "$help_out" "FETCH_COMMENTS_OWNER" verifies the new "Repo resolution" section is reachable via --help. ✓

`fetch-all-pr-comments.test.sh`

Case 9 — Stubs gh repo view to return empty, no env override; asserts exit 2 and that stderr names FETCH_COMMENTS_OWNER. Exercises the exact failure mode the PR fixes. ✓

Case 10 — Sets FETCH_COMMENTS_OWNER=o FETCH_COMMENTS_REPO=r with the same stub; asserts exit 0. Since the stub returns [] for all non-repo view gh calls, the script produces a valid empty JSON array and exits cleanly. The assertion is light (exit 0 only, no output validation) but sufficient to confirm the override unblocks resolution. ✓

Implicit dependency between Cases 9 and 10 — Both share the $TEST_TMPDIR/bin-norepo/gh stub created in Case 9's setup. Sequential shell scripts make this safe, but if tests are ever refactored the dependency is implicit. Low risk for now.

`SKILL.md`

Step 0.1's inline parenthetical is dense (noted as a non-blocking readability nit in the earlier automated review — agreed, not blocking):

fetch-all-pr-comments.sh (derives owner/repo from the current directory; from a cwd that is not a checkout of the target repo, export FETCH_COMMENTS_OWNER/FETCH_COMMENTS_REPO first — also unblocks the readiness gate's exit 4), filter own prior replies, classify

A follow-up breaking the override note into its own sub-bullet would improve scannability. Not blocking.

`reference/loop.md`

§5.0.3 and §5.1.3 additions are accurate and consistent with what the scripts now emit. The §5.0.3 note paraphrases the new error message rather than quoting it verbatim — reasonable given the full message is long. ✓

Other

plugin.json 0.15.7 → 0.15.8 is the correct patch bump for a diagnostics-only change. ✓
CHANGELOG.md entry is thorough, accurate, and follows the existing format. ✓
shellcheck reported clean per PR verification. ✓

claude · 2026-07-21T09:40:10Z

Claude finished @kyle-sexton's task in 1m 44s —— View job

Security Review

Gather PR diff and changed files
Review shell scripts for injection surfaces and unsafe patterns
Review test files for security issues
Check docs/markdown for credential/secret exposure
Post findings

No security issues found.

This PR adds two diagnostic printf lines, updates one existing printf message in fetch-all-pr-comments.sh, adds test cases for both scripts, and updates markdown documentation. Here is what I examined:

$PWD in printf messages (babysit-readiness-gate.sh L152, fetch-all-pr-comments.sh L101)

# babysit-readiness-gate.sh L152 — safe pattern
printf '...current directory (%s) via gh repo view...\n' "$PWD" >&2

# fetch-all-pr-comments.sh L101 — safe pattern
printf '...current directory (%s)...\n' "$PWD" >&2

$PWD is passed as a positional %s argument, not interpolated into the format string — no format-string injection surface. A crafted directory name containing ANSI control sequences could corrupt terminal stderr output, but this carries no code-execution or privilege-escalation vector. Both changes are stderr-only diagnostic messages.

FETCH_COMMENTS_OWNER/FETCH_COMMENTS_REPO → gh api URL construction — The env-var override is used to build gh api "repos/$OWNER/$REPO/..." paths without stripping / or ... This pre-existing pattern (untouched by this PR) could theoretically let a caller direct gh api to unintended GitHub API paths — but gh is constrained to the GitHub API and requires the caller's own auth token, so the blast radius is: read-only access to GitHub API endpoints the caller already has access to. This PR documents the override as the escape hatch; it does not change the code that uses it.

Test stubs (babysit-readiness-gate.test.sh L406, fetch-all-pr-comments.test.sh L200) — written via single-quoted heredocs into $TEST_TMPDIR and prepended to PATH for individual invocations only. Standard safe test-isolation pattern; no variable expansion in the stub body.

PR_NUMBER in diagnostic printf (gate L151) — passed as %s positional argument, not interpolated. Safe.

Documentation changes (SKILL.md, reference/loop.md, CHANGELOG.md) — no executable surface.

usage() line-range update (gate L78) — sed -n '2,65p' "${BASH_SOURCE[0]}" reads from the script's own source file; not user-controlled input.

…de-ops, work-items Re-checked open PRs touching these plugins' plugin.json right before the version-bump collision protocol requires it: source-control carries #839 (0.15.8) and #840 (0.15.9), claude-ops carries #844 (0.17.2), work-items carries #826 (0.18.2) — all still open. Bumps this PR's claims one past each plugin's current highest open-PR claim (source-control 0.15.10, claude-ops 0.17.3, work-items 0.18.3) so this PR does not collide at merge time regardless of which sibling lands first. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01K1V3gkrfSf75isB8MiDy3o

kyle-sexton · 2026-07-21T16:07:04Z

Tower merge-sit: merge-drive lane dark ~5.5h (rate-limit hold since ~10:36Z); tower sitting merges per stall protocol. Gate-verified live: CLEAN, 0 unresolved threads. This was generated by AI (control tower, session 6).

Resolves conflicts on source-control and work-items plugin.json/CHANGELOG.md: main advanced source-control to 0.15.8 (#839, merged) and work-items to 0.18.2 (#826, merged) since this branch was last rebased. Also re-checked live open PRs at merge time and found work-items now carries a new open claim, #857 at 0.19.0 (Jira adapter) — re-bumped this branch's work-items claim from 0.18.3 to 0.19.1 to stay ahead of it. source-control's 0.15.10 and claude-ops's 0.17.3 remain valid (still one past #840/#860's 0.15.9 and #844/#860's 0.17.2 open claims, respectively). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01K1V3gkrfSf75isB8MiDy3o

Integrate #839 (source-control 0.15.8, babysit-readiness-gate diagnostic). Resolved plugin.json to 0.15.9 (one past main) and stacked the 0.15.9 CHANGELOG entry above the merged-in 0.15.8 entry. SKILL.md auto-merged (distinct regions); babysit-prs skill-quality gate green at 499/500 lines.

@me

…gs for babysit snapshot (#882) ## Summary Fixes #511. The babysit PR-queue snapshot derived `self_logins` — the posting identities whose comments self-classification suppresses — from the `--author` discovery filter, conflating two distinct concerns (*which authors' PRs to discover* vs *whose comments to suppress*). That broke in both directions: - **Under-inclusion:** configured extra self identities (`babysit_self_logins`) rode on `--author`, so when autopilot widening drops `--author` entirely, a bot poster's own comments re-fired `new_human_blocking_feedback` every cycle. - **Over-inclusion:** a `--queue --author alice` run authenticated as a different login made `alice` a self-login, suppressing Alice's genuine feedback from the worker-dispatch arm. ## The fix `pr_queue_snapshot.py` now resolves self-identity from dedicated **`--self`** (full override — exactly the given logins, `@me` not added) / **`--extra-self`** (added on top of the authenticated `@me`) flags, mirroring the existing `babysit-readiness-gate.sh` semantics, resolved **independently of `--author`** and before the `--pr`/`--queue` scope split (so single-PR scope is covered too). This supersedes the `@me`-only union added in #494, and the author-derived self fallback in `build_config` is removed so **no discovery author can leak into the self set** in any path. The skill's step-4 invocation and the `babysit_self_logins` userConfig table row now route the configured extras through `--extra-self` instead of overloading `--author @me,<self-logins>`; because self-identity no longer rides on `--author`, it survives autopilot widening. ## Tests (red-green, with transparency on the rewritten spec) - Added `SelfIdentityDecouplingTests` asserting both directions at the `build_snapshot` level (extra-self survives a dropped `--author`; a discovery author is not treated as self). **Verified these fail against the pre-fix author-union code and pass after the fix.** - The `resolve_self_logins` unit tests pinned the *old* author-union contract; they are rewritten to the new flag-based contract, and `test_discovery_authors_are_unioned_with_the_self_login` (which asserted the over-inclusion behavior now fixed) is deleted. `test_raw_author_fallback_drops_me_when_unresolved` becomes `test_missing_resolved_self_logins_yields_empty_not_author`. These are a ratified spec change (the issue's converged decision), not test weakening. ## Verification (local gates) - `python -m unittest discover -s tests -p 'test_*.py'` (the CI entry via `engine.test.sh`) — **341 tests OK**. - `check-changelog-parity.sh --check-bump origin/main` — PASS (0.16.0 entry present). - `check-skill-portability.sh origin/main` — PASS. - `check-changed-skills.sh` — SKILL.md 499/500 lines (under the hard cap); the one reported "engine.test.sh failed" is the pre-existing repo-wide ruff E402 quirk that does **not** reproduce in CI (main is green; merged babysit PRs #860/#839 pass skill-quality-gate), and this change adds zero new ruff errors (32 before == 32 after). - `markdownlint-cli2` on the changed docs — clean; `plugin.json` valid. - Independent fresh-context code review (producer ≠ reviewer). ## Version Per-plugin bump **0.15.9 → 0.16.0** (minor — new `--self`/`--extra-self` CLI surface alongside the behavior fix). `marketplace.json` pins no version for `source-control`, so only `plugin.json` changes. Closes #511 ## Related - #497 — single-`--pr` scope leaves `self_logins` empty (sibling facet of the same seam, triaged blocked-by #511). This change resolves self-identity for `--pr` scope too, but its acceptance is left for the #497 lane to verify — not closed here. - #494 — added the `@me`-only `--queue` union this supersedes. - #473 — engine-level self-filter (related self-dispatch symptom). - #881 — follow-up (filed): the snapshot raises on an unresolvable `@me` where the readiness gate degrades to the `--extra-self` set; tracks whether fail-loud is the intended contract. Surfaced in this PR's review. 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com>

…ot bare name (#484) (#840) ## Summary The `source-control:babysit-prs` worker/autopilot contract told workers to run the merge gate and resolve review threads by invoking the guarded mutation wrappers as **bare command names** (`source-control-babysit-merge …`, `source-control-babysit-resolve-thread …`). Those bare names are **not on the Bash tool's `PATH`**, so every such invocation failed `command not found` (exit 127) — reproduced first-hand in this environment. Workers fell back to raw `gh api graphql resolveReviewThread` calls, losing the wrapper's `--allowed-owners` guardrail and the JSON `action` receipt the contract tells them to parse, and re-implementing the safety logic ad hoc per worker. ## Fix Invoke each wrapper **by its bundled path** — `bash "${CLAUDE_PLUGIN_ROOT}/bin/<wrapper>" …` — exactly the form the read-only sibling scripts under `${CLAUDE_PLUGIN_ROOT}/scripts/` (`fetch-all-pr-comments.sh`, `babysit-readiness-gate.sh`) already use. Updated surfaces: - **`skills/babysit-prs/reference/safety.md`** — the single home for the exact commands. The prior *Guarded Mutation Wrappers* posture said the wrappers are invoked "only by their pinned bare wrapper names … never through an interpreter-prefixed path," which the bare-name PATH gap made unfollowable. It now prescribes the `bin/`-path form and narrows the prohibition to what actually protects the guards: **raw-Python invocation** (`python … babysit_merge.py`, which bypasses the wrapper's own guards such as the merge wrapper's `--allow-unpinned-head` rejection) and **piping a wrapper into an interpreter**. All fenced command blocks and the Pinned-Command Degradation operator handoffs use the path form. - **`skills/babysit-prs/SKILL.md`** — corrected the "invoked ONLY by their bare wrapper names" statement; points at `safety.md` as the single home and states that every `source-control-babysit-<x> …` command in this skill and its references is launched by the `bin/`-path form. - **`skills/babysit-prs/reference/orchestration.md`** — the worker prompt template (the literal contract handed to a dispatched worker) and the file's two other runnable references now use the path form, so the file is self-consistent for a standalone worker. - **`reference/review-discipline.md`** — a one-line pointer in D7.5 noting the babysit tiers resolve through the guarded wrapper (which adds `--allowed-owners`, bot-vs-human classification, and a JSON receipt). D7.5's general raw-GraphQL policy is **unchanged** — it stays the mechanism for `/pull-request`, which is not a babysit-tier consumer; repointing it at the babysit-tier wrapper would wrongly couple `/pull-request` to a babysit tool. Why the path form and not shipping the bare name onto PATH: the wrappers **are** designed to be bare commands on PATH, and Claude Code's plugin reference documents a plugin's `bin/` as added to the Bash tool's PATH "while the plugin is enabled." Empirically that is not happening in this environment — an upstream/harness matter this repo cannot fix. The `${CLAUDE_PLUGIN_ROOT}/bin/`-path form is the deterministic, repo-level invocation that works today and matches the established sibling-script convention. ## Verification - **Reproduced the gap** in this worktree's Bash tool (plugin enabled, Windows/Git-Bash): `which source-control-babysit-resolve-thread` → not found (exit 1); `$CLAUDE_PLUGIN_ROOT` is empty in the shell; no plugin `bin/` on `PATH`. - **Confirmed the path form keeps every guard** (so the posture refinement is safe): invoking a wrapper by path still runs the wrapper — - `bash "…/bin/source-control-babysit-merge" owner/repo#1 --allow-unpinned-head` → still rejected, **exit 2** (`--allow-unpinned-head is not permitted through the wrapper`). - `bash "…/bin/source-control-babysit-resolve-thread" owner/repo#1 --resolve` (no `--allowed-owners`) → still **fail-closed, exit 3**, with the JSON refusal receipt. - **Lint**: `markdownlint-cli2` clean (0 errors) on all four edited docs. - Doc-only change (`.md` + `plugin.json` version); no scripts or executables modified. Closes #484 ## Related - **#484** — this issue. - **#475 / #839** — same bare-wrapper-invocation class (the `babysit-readiness-gate.sh` owner/repo-derivation diagnostic), landing concurrently in `plugins/source-control/`. Per the coordinate-and-serialize protocol this PR is opened **DRAFT + `do-not-merge`**, with its version set **one increment past #839** (`0.15.9` vs #839's `0.15.8`); once #839 lands, drop `do-not-merge` and confirm the version still leads `main`. - **#512** — a sibling open `source-control:babysit-prs` issue (resolve-thread `humanThreadsActed` mislabel), noted as concurrent source-control work. - **Follow-up residual — allow-rule match (not fixed here):** the `${CLAUDE_PLUGIN_ROOT}/bin/…` form does not match a pre-approved bare-name `Bash(source-control-babysit-resolve-thread:*)` allow rule, so an operator's narrow allowlist entries no longer auto-approve these calls (per-call prompts return). The wrappers' own header comments and this repo's `bin/` design still assume the bare-name form; restoring it is deferred to the root-cause fix below. - **Follow-up residual — root cause, tracked as #843:** Claude Code documents a plugin's `bin/` as on the Bash tool's PATH while enabled, but it is empirically absent in this environment. Only closing that upstream/harness gap makes the bare names — and the allow-rule match above — work as intended. (Not fixed here.) --------- Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>

kyle-sexton added the automated Opened by automation. label Jul 21, 2026

kyle-sexton marked this pull request as ready for review July 21, 2026 09:39

kyle-sexton merged commit 3157f44 into main Jul 21, 2026
26 of 27 checks passed

kyle-sexton deleted the fix/475-readiness-gate-owner-repo-derivation branch July 21, 2026 16:07

kyle-sexton mentioned this pull request Jul 21, 2026

control-tower: closure-audit batch 2026-07-20 — operator decisions (fold-member consolidation, re-baseline additions, label normalization, title convention) #607

Open

kyle-sexton mentioned this pull request Jul 21, 2026

feat(source-control): dedicated --self/--extra-self self-identity flags for babysit snapshot #882

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix(source-control): actionable owner/repo diagnostic on readiness-gate fetch failure#839

fix(source-control): actionable owner/repo diagnostic on readiness-gate fetch failure#839
kyle-sexton merged 1 commit into
mainfrom
fix/475-readiness-gate-owner-repo-derivation

kyle-sexton commented Jul 21, 2026

Uh oh!

claude Bot commented Jul 21, 2026 •

edited

Loading

Uh oh!

claude Bot commented Jul 21, 2026 •

edited

Loading

Uh oh!

kyle-sexton commented Jul 21, 2026

Uh oh!

chatgpt-codex-connector Bot commented Jul 21, 2026

Uh oh!

claude Bot commented Jul 21, 2026 •

edited

Loading

Uh oh!

claude Bot commented Jul 21, 2026 •

edited

Loading

Uh oh!

kyle-sexton commented Jul 21, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Conversation

kyle-sexton commented Jul 21, 2026

Summary

Fix

Verification

Related

Uh oh!

claude Bot commented Jul 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Code Review

babysit-readiness-gate.sh

fetch-all-pr-comments.sh

babysit-readiness-gate.test.sh

fetch-all-pr-comments.test.sh

SKILL.md — Minor readability nit

reference/loop.md

Other

Uh oh!

claude Bot commented Jul 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Security Review

Uh oh!

kyle-sexton commented Jul 21, 2026

Uh oh!

chatgpt-codex-connector Bot commented Jul 21, 2026

Uh oh!

claude Bot commented Jul 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Code Review

babysit-readiness-gate.sh

fetch-all-pr-comments.sh

babysit-readiness-gate.test.sh

fetch-all-pr-comments.test.sh

SKILL.md

reference/loop.md

Other

Uh oh!

claude Bot commented Jul 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Security Review

Uh oh!

kyle-sexton commented Jul 21, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

claude Bot commented Jul 21, 2026 •

edited

Loading

`babysit-readiness-gate.sh`

`fetch-all-pr-comments.sh`

`babysit-readiness-gate.test.sh`

`fetch-all-pr-comments.test.sh`

`SKILL.md` — Minor readability nit

`reference/loop.md`

claude Bot commented Jul 21, 2026 •

edited

Loading

claude Bot commented Jul 21, 2026 •

edited

Loading

`babysit-readiness-gate.sh`

`fetch-all-pr-comments.sh`

`babysit-readiness-gate.test.sh`

`fetch-all-pr-comments.test.sh`

`SKILL.md`

`reference/loop.md`

claude Bot commented Jul 21, 2026 •

edited

Loading