Content
86%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 tight, fully executable skill body: complete interface documentation, runnable examples, and verified bundle structure, with only a duplicated 'prefer targeted calls' line and absent error-recovery guidance as minor blemishes.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean bullet-style sections with no concept over-explanation, but "Prefer targeted metadata paths instead of broad unfiltered pulls" (Operating rules) near-duplicates "Prefer targeted endpoint calls instead of broad unfiltered dumps" (Execution behavior) — a minor instance that could be trimmed, so anchor 4 rather than 5. | 4 / 5 |
Actionability | Fully copy-paste-ready: a complete executable bash example with the full JSON payload, three concrete common-case request patterns (genome/taxon, genome/accession, taxonomy/taxon), and exact input, output, and error schemas — all verified to match the actual script interface. | 5 / 5 |
Workflow Clarity | The single action (construct JSON input, pipe to scripts/ncbi_datasets.py, interpret output) is unambiguous, and the failure checkpoint is documented ("Failure returns ok=false with error.code and error.message"), but there is no error-recovery guidance on what to do after a failure, matching anchor 4's minor validation gap. | 4 / 5 |
Progressive Disclosure | Under 50 lines with clearly organized sections; the only referenced bundle file (scripts/ncbi_datasets.py) exists and matches the documented interface, and the body explicitly states no further references are needed — satisfying the simple-skill exception for a 5. | 5 / 5 |
Total | 18 / 20 Passed |