Content
61%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's command reference is excellent and lean, and the working rules encode genuinely non-obvious operational policy with validation checkpoints. Its weaknesses are stylistic: dense prose paragraphs hide sequenced workflows and their verification steps, and specialized procedures are inlined where splitting them out would keep the overview leaner.
Suggestions
Convert the PR-update and fork-reuse rules into numbered step sequences that include the exact executable commands (e.g., gh pr view 123 --repo owner/repo --json headRepository,headRefName,headRefOid) so the 'read back the PR head' verification checkpoint is actionable rather than described.
Move the 'Contributor scope and permission advice' subsection to a separate reference file (e.g., references/contributor-assessment.md) and link it from SKILL.md, keeping the main file focused on gh usage and trimming the dense prose.
Rewrite the remaining 'Working rules' paragraphs as short bullet checklists to cut sentence density while preserving each rule's concrete field names and validation requirements.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The 'Common commands' sections are lean, but 'Working rules' and its subsections pack policy into dense compound sentences (e.g., 'A local upstream or an identically named branch in the base repository does not identify the PR source') that could be tightened into short checklists, and the contributor-scope section is tangential to the skill's stated read-oriented purpose. | 3 / 5 |
Actionability | 'Common commands' provides copy-paste executable gh commands for PRs, CI runs, issues, and API queries, and the rules cite concrete fields ('headRepository', 'headRefName', 'headRefOid', 'permissions.push'). The advanced PR-update and fork-reuse workflows, however, describe steps in prose without the corresponding executable commands, keeping this below a 5. | 4 / 5 |
Workflow Clarity | Validation checkpoints are present ('read back the PR head equal to the intended local commit before claiming an update or merging', 'read back uncertain effects before retrying'), so the destructive-operation cap is not triggered, but the multi-step workflows are embedded in narrative paragraphs rather than clearly ordered steps, leaving the sequence implicit as in the 3 anchor. | 3 / 5 |
Progressive Disclosure | A single well-organized file with clear section headers ('Use this skill when', 'Working rules', 'Common commands') and no nested references, so navigation is easy. At ~63 lines with specialized procedures (fork-reuse recovery, contributor assessment) inlined that could arguably live in separate reference files, it sits at 'good structure, minor organization gaps' rather than a 5. | 4 / 5 |
Total | 14 / 20 Passed |