Content
86%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 skill body: exact commands, deterministic fallback logic, and a complete output template with well-organized sections. The only improvements are removing the repeated skip-condition restatements and adding a lightweight self-check on the emitted PR block.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and imperative, adding only non-obvious notes (e.g. untracked files because "`--stat` omits them", never using a feature branch's upstream as the PR base), but the skip conditions are restated in the Close-out Gate, the Workflow, and the Output Format intro, which could be trimmed. | 4 / 5 |
Actionability | Every step is executable: exact git commands with flags, a deterministic BASE_REF fallback chain (upstream/main → origin/main → main) with merge-base computation, concrete branch-naming patterns, and a copy-paste output template covering the common cases. | 5 / 5 |
Workflow Clarity | The nine steps are clearly sequenced with real checkpoints (compute BASE_REF/BASE_COMMIT first for reuse, empty-combined-diff early exit, "Do not block if the issue cannot be fetched"), but there is no validation of the emitted block itself and the change-type classification steps rely on judgment, leaving minor gaps versus the explicit-feedback-loop anchor. | 4 / 5 |
Progressive Disclosure | The ~60-line body has well-organized sections (Purpose, Close-out Gate, Inputs, Workflow, Output Format), no bundle files are needed, and nothing is inlined that clearly belongs in a separate file, so the structure fully serves navigation. | 5 / 5 |
Total | 18 / 20 Passed |