CI workflow correctness

All CI workflows must be syntactically valid, reliably triggered, and logically non-redundant. Apply these rules: 1) Every step must be valid: each `- name:` entry must include either `uses:` or `run:` (never both missing).

copy reviewer prompt

Prompt

Reviewer Prompt
Deploy Agent

All CI workflows must be syntactically valid, reliably triggered, and logically non-redundant.

Apply these rules: 1) Every step must be valid: each - name: entry must include either uses: or run: (never both missing). 2) Condition + command belong together: put if: on the same step as the run: (or uses:). 3) Triggers must match the repo: ensure the referenced branches in on: push / on: pull_request actually exist; avoid paths-ignore patterns that skip directories containing code. 4) Version pinning clarity: when pinning a GitHub Action to a commit SHA, the inline comment must match the intended release version. 5) Avoid duplicate work: don’t rerun linters/tests already covered by dedicated lint/test workflows or pre-commit jobs. 6) Respect project runtime policy: don’t hardcode an older Python version if the project expects testing against the latest production CPython; align job installs with the intended uv dependency groups.

Example (valid conditional step):

- name: Build docs artifact
  if: $
  run: |
    scripts/build_directory_md.py 2>&1 | tee DIRECTORY.md

Example (pin comment matches intent):

- name: Set up uv
  uses: astral-sh/setup-uv@37802adc94f370d6bfd71619e3f0bf239e1f3b78 # v7.6.0

Example (sanity on triggers):

  • Only ignore files/directories if they truly contain no executable/code paths required for CI, and only reference branches that exist in the repository.

Source discussions