Content
71%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-structured, actionable workflow for an instruction-only skill: concrete checklists, a mandatory output template, and exemplary step-mapped one-level-deep references that all exist on disk. The main weakness is redundancy — Hard Rules, What This Skill Should Not Do, and Quality Standard substantially restate the Execution steps and Task section.
Suggestions
Consolidate Hard Rules with the overlapping Execution step content and the "What This Skill Should Not Do"/"Quality Standard" sections into a single deduplicated rules section, cutting the body substantially.
Add a short worked example (a 3-paper set with a brief sample ranking and citation-role assignment) to move actionability toward fully concrete coverage of common cases.
Add an explicit final validation loop, e.g., "Re-check the completed output against the Hard Rules and Section J before delivering; fix and re-check any violations."
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | There is no padding explaining concepts Claude already knows, but the same rules are restated across Task, Core Function, Execution Steps 2–5, Hard Rules, "What This Skill Should Not Do", and Quality Standard (e.g., design-label vs. evidence value appears at least four times), and the "This skill should/must..." framing repeats heavily. Anchor 3 fits: mostly efficient but could be consolidated. Not 2: each section still carries operational content, not filler. | 3 / 5 |
Actionability | Concrete, executable instruction-only guidance: explicit design-family checklist (Step 2), eight named methodological-quality review dimensions (Step 3), enumerated overclaim patterns (Step 5), five defined citation roles, and a mandatory A–J output structure. Not 5: no worked example (e.g., a sample ranked 3-paper set) is provided, and citation-role definitions are brief. | 4 / 5 |
Workflow Clarity | A clear 8-step sequence with input validation (including a scripted out-of-scope response), a mandatory output template, and an explicit limitations step (Step 8). Not 5: there is no feedback/re-check loop (e.g., re-validate the drafted ranking against the Hard Rules before finalizing), and checkpoint enforcement relies on the reader. | 4 / 5 |
Progressive Disclosure | All nine referenced files exist in references/ and each is mapped to a specific workflow step (e.g., "references/evidence-family-taxonomy.md → use when identifying study design family in Step 2"), one level deep, with SKILL.md acting as a concise orchestrating overview. Not 4: references are explicitly signaled, step-mapped, and complete — no organization gaps found. | 5 / 5 |
Total | 16 / 20 Passed |