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.
Highly actionable content with executable, output-annotated examples and well-sequenced lookup workflows, but the single-file structure inlines an API table and a general microbiome interpretation guide that belong in reference files, and several sections duplicate each other or re-teach concepts Claude already knows.
Suggestions
Move the Core Tools table and the Reasoning Framework (diversity metrics, evidence grading, interpretation guidance) into references/ files (e.g., references/tools.md, references/interpretation.md), keeping SKILL.md as a lean overview with one-level-deep pointers.
Cut the duplicated Tips section (accession prefixes and biome-advice already appear in Key Identifiers and Biome Hierarchy) and remove the Quick Start steps that Workflow 2 repeats verbatim.
Trim general-domain explanations (what Shannon/UniFrac/PERMANOVA measure, what dysbiosis is) down to the tool-calibrated thresholds and cautions that Claude could not supply from its own knowledge.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The Reasoning Framework explains standard concepts Claude already knows (Shannon index, Bray-Curtis/UniFrac, PERMANOVA, Firmicutes/Bacteroidetes ratio, what dysbiosis means), and there is duplication: the Tips section repeats the MGYS/MGYA/MGYG accession prefixes already stated in Key Identifiers, the 'use MGnify_list_biomes first' advice already in the Biome Hierarchy section, and Quick Start overlaps Workflow 2. Mostly efficient, but noticeable trimming is possible. | 3 / 5 |
Actionability | All code is fully executable `tu.run_one_function` calls with real arguments and accessions, expected outputs shown inline (e.g., BacDive strain IDs, the MediaDive medium_id 1611 cross-reference, GMrepo sample counts), plus concrete parameter guidance like the ENA query format and the single-keyword MGnify timeout warning. Copy-paste ready and covering the common cases. | 5 / 5 |
Workflow Clarity | Six clearly sequenced workflows with concrete chained calls, a data-type decision tree, and a pre-flight check ('determine which data type the user has via MGnify_get_study_detail'). Workflows are read-only lookups so the destructive-operation cap does not apply, but there are no error-recovery loops (e.g., what to do when a search returns nothing or an ID doesn't resolve), keeping it below the top anchor. | 4 / 5 |
Progressive Disclosure | No bundle files exist, so the skill is a 320-line monolith: a 25-row tool table that is essentially an inline API reference and ~80 lines of general interpretation guidance (diversity metrics, evidence grading) that clearly belong in reference files. Section headers and navigation within the file are good, but content that should be separate is inline and no references are used at all. | 3 / 5 |
Total | 15 / 20 Passed |