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 highly actionable, well-structured skill body: every core operation has executable examples, five realistic workflows show multi-step usage, and both bundle files referenced are real and one level deep. The main weaknesses are length — much of the per-operation parameter/output detail duplicates the existing reference file and belongs there — and a couple of accuracy slips in the examples, including an import from a nonexistent `scripts.string_enrichment` module.
Suggestions
Fix the enrichment example to import from `scripts.string_api` (the `string_enrichment` function lives there; `scripts/string_enrichment.py` does not exist), and make Workflow 1's Step 6 executable rather than a stub comment.
Move the per-operation parameter/output-column tables, the species table, and the confidence-evidence explanation into `references/string_reference.md`, keeping SKILL.md to an overview with one compact example per operation plus the workflows.
Add explicit validation checkpoints in workflows (e.g., after `string_map_ids`, verify identifiers resolved and check for the documented "Error:" prefix before running network or enrichment queries).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~525-line body contains substantial reference-grade material — full parameter tables, output-column listings, a nine-row species table, evidence-channel explanations, and a troubleshooting section — much of which duplicates content in the existing 455-line `references/string_reference.md`. Content is concrete rather than padded with concepts Claude already knows, but a significant fraction of detail (per-operation parameter/output tables, species table, network-type prose) should live in the reference file, matching the anchor "mostly efficient but some unnecessary explanation or could be tightened". Not a 4: the volume of deferred-to-reference material goes beyond minor trimming; not a 2 because almost everything written is operationally useful and free of filler. | 3 / 5 |
Actionability | Nearly every operation ships executable, copy-paste-ready Python with realistic arguments (e.g. `string_network(['9606.ENSP00000269305', ...], required_score=700)`) plus concrete output-column and threshold guidance. Not a 5: the enrichment example imports from a nonexistent module (`from scripts.string_enrichment import string_enrichment` — only `scripts/string_api.py` exists), Workflow 1 ends with the non-executable stub "# Step 6: Parse and interpret results", and Workflow 4 uses a placeholder `'gene_name'` identifier. | 4 / 5 |
Workflow Clarity | Five named workflows (protein-list analysis, single-protein investigation, pathway-centric, cross-species, network expansion) are presented as clearly numbered, sequenced steps with a strong "Always map identifiers first" convention and a troubleshooting section covering error recovery. Not a 5: there are no explicit validation checkpoints between steps (e.g., verify the mapping returned identifiers or check for the documented "Error:" prefix before proceeding with network/enrichment queries), so checkpoints are implicit rather than stated. | 4 / 5 |
Progressive Disclosure | The bundle structure is real and well-signaled: `scripts/string_api.py` (verified to exist and contain all eight advertised functions) and `references/string_reference.md` (verified to exist, one level deep), with a "Detailed Reference" section explicitly enumerating what the reference covers. Not a 5: the body itself is ~525 lines — far beyond an overview — with parameter/output detail inlined that belongs in the reference file, so the split between SKILL.md and the reference is imbalanced. | 4 / 5 |
Total | 15 / 20 Passed |