Content
67%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 delivers a well-sequenced, genuinely actionable design workflow with a clearly signaled one-level-deep reference bundle. Its main weakness is redundancy: prohibition and hard-rule content is restated across multiple body sections and largely duplicated in the reference files, inflating token cost without adding guidance value.
Suggestions
Consolidate the overlapping prohibition sections — 'What This Skill Is Not For', 'Hard Rules', and 'What This Skill Should Not Do' repeat matching, biomarker-timing, and OR-vs-RR rules; keep one authoritative rule set and delete the duplicates.
Move the detailed Hard Rules and Formatting Expectations content into the existing reference modules (especially references/05_output-style-and-hard-rules.md) and keep only a brief pointer in SKILL.md, reducing the body to a lean overview.
Add one worked mini-example of the mandatory output structure (e.g., a filled-in Bias-Control Matrix row or Variable Collection tier table) to make the required output format fully concrete.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient domain decision logic rather than basic-concept padding, but prohibitions repeat across sections: matching rules appear in 'What This Skill Is Not For', 'Hard Rules > Matching Discipline', Step 5, and 'What This Skill Should Not Do'; source-population, biomarker-timing, and OR-vs-RR rules similarly recur, and 'Quality Standard' mostly restates earlier content. It matches the score-3 anchor (mostly efficient but could be tightened) rather than score 2, since the bulk is genuinely skill-specific guidance rather than filler or explanations of things Claude already knows. | 3 / 5 |
Actionability | As an instruction-only skill, it provides concrete, executable guidance: a 10-step execution logic with explicit decision options at each step, a 12-section mandatory output structure, and defined input-extraction and non-fabrication rules. It matches the score-4 anchor (mostly executable guidance with concrete specifics, minor gaps) — it stops short of score 5 because it offers no worked example of the required output sections or tables. | 4 / 5 |
Workflow Clarity | The 10-step sequence is clear and ordered, with a design-fit gate in Steps 1–2 ('If the user's actual goal is not well served by a case-control design, say so clearly'), an input-validation pass before output, and a Quality Standard checklist. This matches the score-4 anchor (clear sequence, most checkpoints present, minor validation gaps); it lacks the explicit validate-output-then-revise feedback loop of the score-5 anchor. | 4 / 5 |
Progressive Disclosure | Five real, one-level-deep reference files are clearly signaled, each with a 'Use to...' purpose line and mandated as active reading, and the paths verified to exist. It sits at the score-4 anchor rather than 5 because the body is not a lean overview: it inlines substantial material duplicated in the references — notably 'Hard Rules' and 'Formatting Expectations' overlap heavily with references/05_output-style-and-hard-rules.md, and matching/exposure rules overlap reference 03. | 4 / 5 |
Total | 15 / 20 Passed |