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.
The body is highly actionable with executable setup and search commands plus practical cost guidance and error handling. Its weaknesses are redundancy (duplicated async notes and triplicated examples) and an async polling workflow that is described but never shown as a concrete command.
Suggestions
Show an executable polling command, e.g. curl -s "$GOOSEWORKS_API_BASE/v1/proxy/orthogonal/run" with the GET /person/search?request_id=<id> body, and state how to detect completion.
Consolidate the three near-identical Examples into one plus a note on varying the query string, and state the async polling behavior once instead of in both Response and Error Handling.
Remove or trim the <details> curl equivalent block, which duplicates the primary example without the auth header and adds no new information.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body repeats material: the async-polling note appears twice ("Nyne searches are async" in both Response and Error Handling), the three Examples are near-identical curl commands, and the <details> "curl equivalent" block duplicates the main example (minus auth). Mostly efficient commands, but several sections could be tightened, matching the 'some unnecessary explanation or could be tightened' anchor rather than 'minor instances'. | 3 / 5 |
Actionability | Setup and search commands are fully executable copy-paste curls with concrete query examples ("Sam Altman OpenAI", "Jensen Huang NVIDIA CEO"), but the polling step that every search requires is only described in prose ("Poll with GET /person/search?request_id=<id>") with no executable command, a minor gap that keeps it below fully copy-paste-ready. | 4 / 5 |
Workflow Clarity | The async sequence is described (POST returns a request_id, poll GET until complete) with recovery guidance for 404 (retry with different spellings/company) and 429 (wait and retry), fitting 'clear sequence with most checkpoints present; minor validation gaps'. It is not a 5 because the polling loop is never shown as an explicit validate-and-retry step with a command. | 4 / 5 |
Progressive Disclosure | The skill has no bundle files and its sections (Setup, Usage, Parameters, Response, Examples, Error Handling, Tips) are clearly organized with all content core to the single task — good structure with minor organization gaps. It does not reach 5 because the skill exceeds 50 lines and carries some inline duplication (repeated async notes, triplicated examples) that better structuring or a split would address. | 4 / 5 |
Total | 15 / 20 Passed |