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, actionable bootstrap workflow with concrete gh commands, explicit gating before the save step, and well-organized sections. The only gap is a missing explicit validation pass on the synthesized prompt before saving.
Suggestions
Add an explicit verification checkpoint before save_review_style_prompt (e.g., 'Confirm the prompt names >=5 concrete repo-specific bug patterns and a calibrated do-not-flag list before saving').
State the expected failure/retry behavior when gh returns rate-limit or empty review arrays, so the sparse-batch guidance is complete.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean throughout: assumes Claude knows what gh, PRs, and review feedback are, uses no padding, and every instruction earns its place (e.g. 'skip [bot] accounts and obvious automation like codecov / dependabot'). | 5 / 5 |
Actionability | Provides copy-paste-ready gh CLI commands, a concrete save_review_style_prompt call with named fields, and specific numeric targets (400-1200 words, 2-4 sentences, ~8 comments, ~25+ PRs) covering the common cases. | 5 / 5 |
Workflow Clarity | A clear 3-step Research/Extract/Save sequence with explicit gating ('Only after real research, call save_review_style_prompt once') and retry guidance for sparse batches, but lacks a hard output-validation checkpoint on the synthesized prompt itself. | 4 / 5 |
Progressive Disclosure | A single well-organized file with clearly numbered sections (## 1, 2, 3) and no nested references; under 50 lines with no need for external bundles, so the simple-skill exception applies. | 5 / 5 |
Total | 19 / 20 Passed |