[dclong]

Summary

• feat: add lint_repo script to detect and lint repository languages

Changed files

Added

• github_rest_api/scripts/lint_repo.py (+556/-0)
• tests/test_lint_repo.py (+608/-0)
  Modified
• pyproject.toml (+1/-0)

Commits

• 815e332 feat: add lint_repo script to detect and lint repository languages

[dclong]

Claude finished @dclong's task in 2m 52s —— View job

---

Claude's Review of PR #139

• Gather context (PR diff, changed files, CLAUDE.md)
• Review lint_repo.py and test_lint_repo.py manually (the /code-review:code-review skill wasn't invokable in this sandboxed environment, and most Bash commands beyond read-only git diff required approval that wasn't available, so I could not execute pytest/ruff/ty here — review is based on careful manual reading of the diff)
• Post review feedback
• Decide on auto-merge eligibility

Summary

This adds github_rest_api/scripts/lint_repo.py, a well-structured script that detects a repo's languages from git-tracked files and runs the matching lint checks (mirroring lint.yaml's detect + per-language jobs). It's registered as a uv script entry point in pyproject.toml, and comes with a thorough test suite (tests/test_lint_repo.py, 60+ tests) covering detection, check construction, missing-tool handling, fail-fast semantics, and the CLI.

Findings

Minor (non-blocking): summary count double-counts non-fatal failures as "passed"

In run_checks (github_rest_api/scripts/lint_repo.py:383-430), a check that fails but has fatal=False (e.g. markdown: lychee) is neither appended to failed nor skipped:

elif check.fatal:
    failed.append(check.name)
    print(f"FAIL: {check.name}", flush=True)
    ...
else:
    print(f"WARN (non-fatal): {check.name}", flush=True)

But the final tally is passed = len(ran) - len(failed) - len(skipped), so a non-fatally-failed check gets counted as "passed" in the printed Summary: N/M passed line even though it actually failed (just non-fatally). This doesn't affect the returned failed-checks list or the process exit code — only the human-readable summary text is misleading. Worth a follow-up but not a blocker.

Fix this →

Other observations (no action needed)

• list_tracked_files correctly relies on dulwich's Index.__iter__ yielding tracked path keys, matches the project's dulwich-first convention from CLAUDE.md, and handles non-UTF-8 filenames via surrogateescape (covered by a dedicated test).
• The late-binding closure pitfall in build_python_scripts_checks (looping to build a run=lambda cwd, s=script: ... per script) is correctly avoided via the s=script default argument.
• Missing-tool handling is centralized in run_check so both command- and run-based checks get --on-missing-tool support for free, and a run callable is confirmed (by test) to never execute when a required tool is missing.
• pyproject.toml's new scripts.lint_repo entry is correctly placed in alphabetical order.

No correctness bugs found that would affect actual lint outcomes or exit codes. The one issue above is cosmetic/logging-only.

AUTO_MERGE_APPROVED