Content
88%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 well-crafted process skill: fully actionable commands and sub-agent briefs, a clearly sequenced workflow with fail-early validation and missing-spec fallbacks, and disciplined section structure. The only improvements are marginal — trimming known-concept restatement and considering whether the smell baseline belongs in a reference file.
Suggestions
Trim the 'Why two axes' section to one or two lines and drop the parenthetical restatement of the two axes at the top, since the description and Process section already establish them.
Consider moving the 12-smell baseline to a references/ file and having the Standards sub-agent read it, keeping SKILL.md as a shorter overview — or explicitly state that sub-agents cannot reliably read bundle files to further justify the inline placement.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The process text is lean with concrete commands and verbatim sub-agent briefs, and the 12-smell baseline is deliberately inlined so the sub-agent prompt can carry it ("sub-agent 没有其他方式读取它"). It is not a 5 because the smell catalog and the 'Why two axes' section restate material Claude largely knows and could be tightened. | 4 / 5 |
Actionability | Fully executable guidance: exact commands (`git diff <fixed-point>...HEAD`, `git log <fixed-point>..HEAD --oneline`, `git rev-parse <fixed-point>`), complete quoted sub-agent briefs with word limits, and a concrete fallback order for locating the spec. Not a 4 because no key detail is missing for the common cases. | 5 / 5 |
Workflow Clarity | A clear five-step sequence with explicit validation in step 1 ("确认 fixed point 能解析…并且 diff 非空…应该在这里失败，而不是…后才失败") and error recovery paths (ask the user for a spec, skip the Spec sub-agent and report "no spec available"), plus explicit aggregation rules. This is a read-only review skill, so the destructive/batch validation cap does not apply. | 5 / 5 |
Progressive Disclosure | The single SKILL.md (no references/, scripts/, or assets/ exist) is well-sectioned and self-contained, with the only external pointer being the repo's `docs/agents/issue-tracker.md`. It is not a 5 because the ~20-line smell baseline is bulk reference material kept inline and the body exceeds the under-50-line simple-skill exception; it is not a 3 because the structure is clean and the inline choice is explicitly justified. | 4 / 5 |
Total | 18 / 20 Passed |