Content
81%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is a tightly written, highly executable PR review workflow: concrete gh/git/python commands, explicit validation and error-recovery loops, conditional steps, and a clean trigger-gated reference pattern for special cases. The main gaps are the missing referenced bundle file (identifier-normalization.md) and a few spots that could be more precise or trimmed.
Suggestions
Ensure identifier-normalization.md actually ships with the skill bundle — the body links to it but no such file exists alongside SKILL.md, so the triggered read would fail at runtime.
Tighten step 3: the bullet list explaining main/optional/dev dependency categories largely duplicates what the script output already shows; keep only the presentation rules (WARNING blocks, exit-code handling).
Make the CI check matching in step 4 concrete — specify the exact check name pattern or a gh pr checks --json filter instead of the fuzzy 'lint on all python versions / lint*'.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and imperative throughout ('gh pr view PR_ID --json title,body,...', 'git fetch origin devel') and assumes Claude's competence, with only minor trimmable material such as step 3's bullet list explaining dependency categories that partially duplicates the script's own output. Not 5 because a few spots could be tightened. | 4 / 5 |
Actionability | Concrete, executable commands throughout — exact gh invocations with defined JSON fields, the dependency-check script call, and log-fetch commands — with PR_ID substitution clearly defined. Minor gaps keep it below 5: the lint-check match pattern 'lint on all python versions / lint*' is fuzzy, and analysis steps rely on direction rather than commands where commands can't apply. | 4 / 5 |
Workflow Clarity | Clear numbered 1-9 sequence with explicit validation steps and feedback loops: step 1 verifies cwd with pwd, retries with cd, and stops on failure; step 3 branches on script exit codes 1 and 2; step 4 fetches failed-check logs; step 7a requires producing a checklist. Conditional skips are explicit ('If pyproject.toml is not in the diff, skip this step'). | 5 / 5 |
Progressive Disclosure | Good structure with a well-signaled, one-level-deep, trigger-gated reference ('read the linked file ONLY when the PR's diff matches the trigger') for identifier-normalization.md. However, that referenced file is not present in the skill bundle (no references/, scripts/, or assets/ directories exist), so the link cannot be verified to resolve — a navigation gap that keeps it below 5. | 4 / 5 |
Total | 17 / 20 Passed |