Content
85%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A well-structured, highly actionable Methods-writing skill with a concrete IMRAD algorithm, detailed reproducibility and statistical guidance, a worked example, and correctly signaled one-level-deep reference files. The main weakness is token efficiency: several generic process-hygiene sections restate what Claude already knows and could be trimmed.
Suggestions
Trim or collapse the generic boilerplate sections ('Required Inputs', 'Recommended Workflow', 'Validation and Safety Rules', 'Failure Handling', 'Input Validation', 'Quick Validation', 'User Checkpoints') into a few concise lines, since they restate process hygiene Claude already applies.
Move the long worked example draft into a reference file (e.g., references/methods_example.md) and keep only a short pointer plus the most instructive fragment inline, improving token efficiency while preserving actionability.
Tie the validation/checkpoint language to Methods-section-specific failure modes (e.g., missing reporting-checklist items, mismatched Methods-vs-Results content) rather than generic scope/validation phrasing.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The IMRAD algorithm, statistical-reporting parameters, terminology rules, and worked example are substantive and skill-specific, but generic boilerplate sections ('Required Inputs', 'Recommended Workflow', 'Validation and Safety Rules', 'Failure Handling', 'Input Validation', 'Quick Validation', 'User Checkpoints') restate process hygiene Claude already applies, fitting 'mostly efficient but includes some unnecessary explanation'. | 2 / 3 |
Actionability | Provides a concrete 10-step Methods-section algorithm with itemized specifics (e.g., 'italicize gene symbols (e.g., *TP53*)', 'Cohen's d, η², OR, RR and 95%/99% CIs', 'block randomization') plus a full manuscript-ready example draft, matching 'concrete, executable guidance' for an instruction-only skill. | 3 / 3 |
Workflow Clarity | The numbered IMRAD-aligned algorithm gives a clear sequence, supplemented by an explicit final validation pass ('Run a final validation pass for completeness, consistency, and safety') and user-checkpoint guidance; this is a generation skill rather than a destructive/batch operation, so the sequence with checkpoints fits the top anchor. | 3 / 3 |
Progressive Disclosure | Three real reference files (imrad_structure.md, reporting_guidelines.md, writing_principles.md) are referenced one level deep and signaled inline at the relevant steps ('if available'), plus a concise optional reference list, matching 'clear overview with well-signaled one-level-deep references'. | 3 / 3 |
Total | 11 / 12 Passed |