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.
The body is an actionable, well-sequenced instruction skill with explicit validation and feedback loops, well-organized into clear sections. Its main weakness is redundancy across the repeated 'never silence' sections.
Suggestions
Consolidate the three overlapping 'never silence a check' sections (the cause/silence framing, the 'You may NEVER silence a check' list, and 'Hard nevers') into one authoritative list to cut repetition.
Tighten the intro paragraphs that restate the same 'fix the cause, not the check' thesis twice before the workflow begins.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly lean and assumes Claude's competence (e.g. concrete triage categories, named suppression directives), but the 'never silence a check' theme is restated across three sections ('The one thing that matters', 'You may NEVER silence a check', 'Hard nevers') with overlapping lists that could be consolidated. | 4 / 5 |
Actionability | Gives concrete executable guidance — `./scripts/local-checks.sh`, `claude -p "/fix-local-checks <feature>"`, a commit-message example `fix(local-checks): trim query whitespace` — with specific triage buckets; minor gaps are appropriate for an instruction-only skill. | 4 / 5 |
Workflow Clarity | A clearly numbered 5-step sequence with an explicit re-verify checkpoint (step 4: re-run the gate before committing) and a feedback loop (step 5: honest partial progress + STUCK when unfixable), matching the validate→fix→retry pattern the rubric rewards. | 5 / 5 |
Progressive Disclosure | Single well-organized file with clear section headers and no nested references; no bundle files are needed, though at ~100 lines it exceeds the simple-skill 50-line threshold that would justify a 5. | 4 / 5 |
Total | 17 / 20 Passed |