Content
85%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 strong, token-efficient domain skill: concrete tool contracts with real example identifiers, a clearly phased workflow, and genuine error-recovery guidance. Its main structural weakness is that everything lives inline in SKILL.md with no progressive disclosure into reference files, which pushes it toward a monolithic API listing.
Suggestions
Move the per-tool parameter reference (Phases 0–5) into a references/tool-api.md file, keeping SKILL.md as a phased overview with clearly signaled links, so the heavy API detail is loaded on demand.
Consolidate the 'LOOK UP DON'T GUESS' bullets into the phase sections (or vice versa) and drop the opening line that duplicates the frontmatter description, trimming duplicated tokens.
Consider splitting the Common Mistakes and Fallback Strategies tables into a references file or merging fallback guidance into each phase so recovery guidance sits next to the step it recovers from.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense reference material (tool signatures, example IDs, mistake/fallback tables) that assumes Claude's competence, with only minor trim candidates: the opening line duplicates the frontmatter description and the 'LOOK UP DON'T GUESS' bullets restate content already in the phase sections. | 4 / 5 |
Actionability | Every tool is listed with required parameters, literal ID formats ("hsa:5230", "R-HSA-70171", "HMDB0000122"), and exact API constraints ("space-separated string, NOT array", "moverz/REFMET/{mz}/{adduct}/{tolerance}"), giving fully executable lookup guidance for this instruction-only skill. | 5 / 5 |
Workflow Clarity | A clear Phase 0–5 sequence with an overview diagram, an identity-resolution checkpoint ('Always verify metabolite identity first'), and explicit error-recovery loops via Fallback Strategies and Common Mistakes; no destructive/batch operations so the validation cap does not apply. | 5 / 5 |
Progressive Disclosure | The skill is well-sectioned with headers, but it is a single ~150-line file with no bundle files or references at all — the per-tool API reference and the mistake/fallback tables are inlined and could be split into one-level-deep reference files to keep SKILL.md an overview. | 3 / 5 |
Total | 17 / 20 Passed |