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.
Well-structured and actionable with excellent progressive disclosure via real, clearly-mapped reference files. The main weakness is redundancy across Core Function, Execution, Output Structure, and the two "should not" lists, plus the absence of a worked input→output example.
Suggestions
Consolidate overlapping sections: Core Function, Execution Steps 1–9, and Mandatory Output Structure A–J describe the same work three times — merge or cross-reference them to cut repetition.
Remove the duplicated negative guidance: "What This Skill Should Not Do" repeats the earlier "This skill should not" list; keep a single canonical version.
Add one worked input→output example (a vague objective transformed into plain-language, protocol-ready, and downstream-ready forms) to make the actionability concrete and cover the common case.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Individual sections are lean bullet lists and it avoids explaining basic biomedical concepts, but Core Function (10 items), Execution Steps 1–9, Mandatory Output Structure A–J, Hard Rules (14), and two separate "should not" lists substantially overlap and could be consolidated. | 3 / 5 |
Actionability | Concrete, specific instruction-only guidance — explicit audit elements (population, exposure, outcome, comparison, time horizon), named structure types, and a fixed output scaffold — but a worked input→output refinement example is missing, leaving the common case uncovered. | 4 / 5 |
Workflow Clarity | A clear 9-step sequence maps each reference module to its output section, with assessment checkpoints (Step 8 measurability, Section J risk-of-misframing); minor validation gaps and no explicit retry loop keep it just below 5. | 4 / 5 |
Progressive Disclosure | All ten referenced files exist in references/ and are signaled one level deep with explicit usage triggers (e.g. "objective-type-taxonomy.md → use when ... Section B"), giving a clear overview that points cleanly into well-organized detail files. | 5 / 5 |
Total | 16 / 20 Passed |