fix(scripts): derive usage() header ranges instead of hardcoded sed

[kyle-sexton]

Closes #3424

Summary

--help for scripts/affected-tests.sh and datamuse.sh extracted a hardcoded sed -n range that had drifted behind the header, so the printed usage cut off mid-block. The listed babysit scripts already use derived extraction.

Fix

Replace the hardcoded ranges with the babysit-readiness-gate.sh awk (shebang to first non-comment line). datamuse.sh also accepts --help / -h (exit 0). songwriting 1.4.20.

Verification

scripts/affected-tests.sh --run from the worktree: 129 shell suites passed. Exit 3 is the documented NOT RUN for 7 selected Python suites. New assertions pin the last header sentence and the datamuse Examples block.

Related

N/A

  

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-02T08:57:56.675128Z	b1eee28	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[claude]

Claude finished @kyle-sexton's task in 3m 52s —— View job

---

Code review complete

• Gather context (PR diff, CLAUDE.md, skip gate)
• Review changed files for correctness and maintainability
• Adversarial verification of candidate finding
• Post findings (inline comments + summary)

Skip gate: PR is open, not a draft, and has a real review surface (two usage() extraction fixes + tests) — proceeded with full review.

Summary

The core fix is sound: replacing the hardcoded sed -n line ranges in scripts/affected-tests.sh and datamuse.sh with the awk-derived header extraction (shebang → first non-comment line) matches the existing convention in plugins/source-control/scripts/babysit-readiness-gate.sh, and I traced both header blocks line-by-line to confirm the derived range now reaches the intended content (the "Both stages fail loud" sentence in affected-tests.sh, and the full Examples block through LIMIT=50 datamuse.sh near grief in datamuse.sh) without an early exit on a stray blank line. datamuse.sh's new --help/-h handling (exit 0, no request issued) is correctly placed before the mode/arg validation.

One finding, posted inline:

• plugins/songwriting/context/pat-pattison/scripts/datamuse.test.sh:193 — the new --help test block (lines 188–192) was inserted between two assertions that used to both cover the same "no arguments" run_datamuse call. Since run_datamuse overwrites the global $OUT/$RC/$ERR on every invocation, the assertion at line 193 (still labeled "no arguments writes nothing to stdout") now actually checks the --help call's stdout, not the no-args call's. It still passes today (both paths produce empty stdout), so this won't fail CI, but the original "no arguments → empty stdout" invariant is no longer covered by anything, and the label is misleading. Verified independently via a second pass before posting.

Nothing else surfaced — the plugin.json/CHANGELOG.md version bump and changelog wording match the actual line ranges affected, and scripts/affected-tests.test.sh's new end-of-header assertion is consistent with how the suite is always invoked (CWD at repo root).

Branch: cursor/fix-3424-usage-sed-ranges-df93

[claude]

Claude finished @kyle-sexton's task in 58s —— View job

---

Security review

• Gather context (PR diff, changed files)
• Apply skip gate
• Review changed files for security-relevant surface
• Report findings

Skip gate: no security-relevant surface. This PR only changes how usage()/--help text is
extracted in two shell scripts (scripts/affected-tests.sh,
plugins/songwriting/context/pat-pattison/scripts/datamuse.sh), replacing a hardcoded sed -n
line range with an awk scan that reads the script's own header (shebang → first non-comment
line) via ${BASH_SOURCE[0]}. There's no user-controlled input into the new awk/usage() code
path, no auth/access-control logic touched, no secrets/token handling, and no GitHub Actions
workflow or permissions changes — the remaining diff is a version bump, changelog entry, and test
assertions.

No security issues found.

[github-actions]

Last security-reviewed head: b1eee280fadb8189ec9e8180880f179a1bdccdf9. On the next push, the relevance gate compares only the commits since this SHA; delete this comment to force a full re-review.

[github-actions]

Claude has reviewed this PR 2 times. The lane skips further automatic reviews after 5; deleting this comment resets the count.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: b1eee280fa

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[claude]

Claude finished @cursor[bot]'s task in 2m 18s —— View job

---

Code review complete

• Gather context (PR diff, CLAUDE.md, skip gate)
• Review changed files for correctness and maintainability
• Adversarial verification of candidate findings
• Post findings (inline comments + summary)

Skip gate: PR has a real review surface and is receiving pushes (synchronize event) — proceeded with review.

Summary

This push (8bdaf64) contains exactly one change since the last-reviewed commit (b1eee28): datamuse.test.sh.

That change fixes the finding both this lane and Codex flagged on the prior commit: the "no arguments writes nothing to stdout" assertion was mislabeled because it ran after the newly-inserted run_datamuse --help block had already overwritten the shared $OUT/$RC/$ERR globals, so it was silently checking the --help invocation instead of the no-args one.

The fix moves that assertion to line 187, immediately after the no-args run_datamuse call (line 175) and before reset_stub/run_datamuse --help (lines 189–190), and gives the --help block its own correctly-labeled stdout assertion at line 194 ("--help writes nothing to stdout"). Both invariants are now independently covered under the invocation they actually name — I traced the global-overwrite ordering line-by-line to confirm no other assertion in this file reads $OUT/$RC/$ERR across a reset_stub/run_datamuse boundary the way this one used to.

No new findings. The rest of the PR (the awk-derived usage() header extraction in scripts/affected-tests.sh and datamuse.sh, the new --help/-h handling, and the pinned header/Examples-block assertions) was already reviewed on the prior commit and is unchanged here.