Content
81%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 highly actionable, well-sequenced workflow skill: exact globs, grep patterns, verbatim prompts, complete output templates, and explicit guards with named recovery paths for every failure mode. Its weaknesses are token cost from repeating the qa.level guard and cross-skill meta-commentary, and inlined templates that inflate the file to ~330 lines with no bundle files distributing detail.
Suggestions
State the qa.level: minimal stop once in Section 1 and reference it from the intro and Section 7 verdict list instead of re-explaining it in each place.
Trim the zero-denominator blockquote to the operative rule — drop the /scope-check cross-reference and the quoted wording from the other skill.
Move the report and suite-manifest templates to a references/templates.md file and keep only field summaries in SKILL.md.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense project-specific instruction with no teaching of known concepts, but the "qa.level: minimal" stop is explained three times (intro paragraph, Section 1, Section 7 verdicts) and the denominator blockquote spends lines on cross-skill meta-commentary ("This is the same defect /scope-check carries a Phase 4 guard against, in the same words"). Not 4 because the repetition and quoted /scope-check wording exceed minor over-explanation; not 2 because there is no generic padding and nearly every line is operational. | 3 / 5 |
Actionability | Concrete globs ("tests/unit/**/*_test.*"), exact tool usage ("Grep tool (pattern ^#, content mode)" with Read offset/limit), verbatim AskUserQuestion prompt text, and complete report and manifest markdown templates make the guidance copy-paste ready. Placeholders like "[Engine-specific command to run all regression tests]" are justified engine-agnostic flexibility rather than pseudocode, so this sits at the fully-executable anchor rather than the 4 anchor's 'minor gaps'. | 5 / 5 |
Workflow Clarity | A seven-step numbered sequence with explicit validation checkpoints and error recovery: the early guard ("STOP here, before any scan"), the zero-denominator guard ("do not emit a percentage"), the smoke-report stop naming the fix ("run /smoke-check first"), ask-before-write gating, and a first-match-wins verdict list. Not 4 because checkpoints are explicit feedback loops with named recovery actions rather than implicit or partially present. | 5 / 5 |
Progressive Disclosure | Well-organized numbered sections with clearly signaled one-level-deep pointers to project docs (".claude/docs/automation-modes.md", ".claude/docs/workflow-modes.md", ".claude/docs/config-resolution.md"), and no bundle files exist to nest. Not 5 because at ~330 lines the full manifest and report templates are inlined where a references/ file could carry them; not 3 because structure and navigation are clean with no buried references. | 4 / 5 |
Total | 17 / 20 Passed |