Content
93%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.
A tight, highly actionable instruction skill: executable gh commands, quantified thresholds, and a well-gated three-step workflow. The only gap is the absence of an explicit post-save validation step before finishing.
Suggestions
Add a brief verify-before-finish step (e.g., confirm the saved prompt is within the 400–1200 word range and reflects patterns actually observed in the sampled PRs) to complete the feedback loop.
State what to do if gh API calls fail or rate-limit mid-research, mirroring the existing sparse-batch escalation guidance.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~58-line body is lean and assumes Claude's competence: no generic concept explanations, no padding; every line carries instruction ('skip [bot] accounts', 'never run gh auth login', 'Do not call read_finding_outcomes in this mode — it will be empty'). | 5 / 5 |
Actionability | Provides copy-paste-ready gh commands (pr list with exact flags, three gh api endpoints) plus concrete thresholds: at least 8 substantive human comments, top ~5 reviewers, 400–1200 words, 2–4 sentences, ~25+ merged PRs fallback rule. | 5 / 5 |
Workflow Clarity | Clear Research → Extract → Save sequence with explicit gates ('Only after real research', sparse-batch escalation, verify preloaded samples with gh), but there is no post-save verification loop, so it fits 'clear sequence with most checkpoints; minor validation gaps' rather than the full feedback-loop anchor. | 4 / 5 |
Progressive Disclosure | A single-file skill with no bundle files (references/, scripts/, assets/ absent) and no nested references; well-organized numbered sections with no content that belongs in a separate file — meets the simple-skill top anchor. | 5 / 5 |
Total | 19 / 20 Passed |