Content
75%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, information-dense skill body: a clear phase workflow, concrete tool syntax, domain-specific gotchas, fallbacks, and honest limitations. The consistent weakness across dimensions is the last increment — trimmed duplicate framing, one executable example per database tool, explicit validation checkpoints, and offloading detail tables to reference files would lift it.
Suggestions
Remove the introductory paragraph that restates the frontmatter description, and fold the 'Guiding principles' list into the phase notes it duplicates.
Add one copy-paste-ready example call for each core tool (MGnify_search_studies, GMrepo lookup) mirroring the concrete kegg_search_pathway(keyword=...) pattern, plus a minimal pandas/scipy snippet to make the 'COMPUTE, DON'T DESCRIBE' mandate executable.
Move the KEGG pathway-significance table and per-database API quirks into a single one-level-deep reference file (e.g., references/tool-notes.md), keeping SKILL.md as a lean overview; add an explicit step after each phase to sanity-check results (e.g., non-empty result count) before proceeding.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with domain facts Claude cannot reliably recall (tool parameter names, MeSH mappings, MIMAG quality thresholds) and avoids explaining known concepts. It is not a 5 because the opening paragraph restates the frontmatter description and the 'Guiding principles' list overlaps the phase notes, tokens that could be trimmed. | 4 / 5 |
Actionability | Concrete guidance is given with exact call signatures (GTDB_search_taxon(operation="search_taxon", query=name), kegg_search_pathway(keyword=...), study_title="*IBD*"), so most instructions are executable as written. It is not a 5 because no complete executable example exists for MGnify or GMrepo calls, and the mandated Python computation has no sample code, leaving minor gaps. | 4 / 5 |
Workflow Clarity | An explicit 8-phase sequence is paired with per-phase notes and error-recovery fallbacks ('If ENA fails, fall back to MGnify', 'GMrepo 0 results: use formal MeSH terms or NCBI taxon IDs'). It is not a 5 because there are no explicit result-validation checkpoints — nothing instructs verifying result sanity before proceeding to the next phase. | 4 / 5 |
Progressive Disclosure | Sections are well organized and clearly navigable, and no bundle files exist so nothing is buried. It is not a 5 because all ~100 lines are inlined — the KEGG pathway significance table and per-tool API quirks are natural candidates for a one-level-deep reference file that would keep the overview leaner. | 4 / 5 |
Total | 16 / 20 Passed |