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.
A well-structured instruction skill: the execution logic, mandatory output structure, and hard rules give Claude a precise, actionable planning procedure, and the 10 reference modules are real and clearly signaled. The main costs are redundancy — boundary rules repeated across three sections and duplicated in reference files — and the lack of a worked example or explicit feedback loop for route failure.
Suggestions
Consolidate the overlapping boundary content in 'Important Distinctions', 'Hard Rules', and 'What This Skill Should Not Do' into one section and delegate detail to references/10-hard-rules.md, keeping only a one-line pointer inline.
Replace the duplicated Mandatory Output Structure / Formatting Expectations detail with a brief summary plus a pointer to references/09-output-section-rules.md, which already restates these rules.
Add a compact worked mini-example (one disease, one route, a three-layer evidence chain) to make the A–L output structure concrete, and add an explicit feedback loop in Step 3 or Step 8 for when evidence weakens and the primary route must be reconsidered.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly tight, decision-oriented bullets with no filler prose, but boundary rules repeat across 'Important Distinctions', 'Hard Rules', and 'What This Skill Should Not Do' (e.g., 'Never collapse these layers into one conclusion' vs. 'Never collapse disease association, mechanistic plausibility, and therapeutic effect into one sentence'), and the Hard Rules section substantially duplicates references/10-hard-rules.md. This matches anchor 3 — mostly efficient but could be tightened — rather than anchor 4, where only minor instances of over-explanation would remain. | 3 / 5 |
Actionability | For an instruction-only skill the guidance is concrete and executable: an 8-step execution logic with worked selection heuristics ('If the user starts from a pathway or target, prefer a target/mechanism-led route'), an explicit DESeq2/limma decision rule ('count data → DESeq2 (recommended default); non-count normalized data → limma'), and a mandatory A–L output structure. It sits below anchor 5 only because no worked mini-example of a finished blueprint is provided to anchor the output format. | 4 / 5 |
Workflow Clarity | The 8-step sequence is clearly ordered, opens with input validation, marks each evidence layer as necessary/recommended/optional, and closes with a risk-audit step and go/no-go gates — most checkpoints are present. It falls short of anchor 5 because there is no explicit feedback loop for error recovery, e.g., instructions for revisiting route selection when the primary route weakens; it is above anchor 3 because checkpoints like Step 8's fallback review are explicit rather than implicit. | 4 / 5 |
Progressive Disclosure | All 10 referenced files exist in references/, are one level deep, and each is clearly signaled with a specific 'Use to...' purpose annotation — good structure per anchor 4. It does not reach anchor 5 because sizable content inlined in SKILL.md (Hard Rules, Mandatory Output Structure, Formatting Expectations) duplicates the corresponding reference modules, which is more than a minor placement gap; references are clearly signaled, keeping it above anchor 3. | 4 / 5 |
Total | 15 / 20 Passed |