Content
50%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 a broad, well-sectioned API catalog with genuinely useful 'Best for' guidance and a thoughtful Tips section, but its execution is undermined by malformed JSON in many of the most prominent curl examples, heavily duplicated request boilerplate, and unexplained async-polling placeholders. Structurally, everything is inlined into one large file rather than split across reference files.
Suggestions
Fix the malformed request bodies: wrap payloads in "body":{...} and balance the quotes/braces in the Tavily /search and /crawl, Exa /search and /findSimilar, and Linkup examples, and remove the stray 'api show <name>' suffixes and backtick from the Discover More section so every example is copy-paste executable.
Deduplicate the curl boilerplate: state once that all calls go to $GOOSEWORKS_API_BASE/v1/proxy/orthogonal/run with Bearer auth and Content-Type headers, then show only the -d payload per example.
Explain the async polling loop for the Tavily/Exa/Valyu deep-research flows: where the {request_id}/{researchId}/{id} values come from in the create response, a suggested poll interval, and what a completed vs. failed status looks like.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The identical curl boilerplate (endpoint URL plus Authorization and Content-Type headers) is repeated verbatim in roughly 30 code blocks, and the 'Discover More' section is redundant and garbled. There is no concept over-explanation, so it sits above anchor 2, but the heavy duplication could be collapsed with a documented variable or a single stated header convention. | 3 / 5 |
Actionability | Commands are concrete curl calls with real parameters, but many flagship blocks (Tavily /search and /crawl, Exa /search and /findSimilar, both Linkup searches) contain malformed request bodies — the -d string closes early, the "body":{...} wrapper is missing, and quotes are unbalanced — plus stray 'api show exa' tokens and a stray backtick in 'Discover More'. These are not copy-paste executable, fitting anchor 3's 'incomplete; missing key details' better than anchor 4's 'minor gaps'. | 3 / 5 |
Workflow Clarity | Async research flows do show a Step 1 create / Step 2 poll sequence and the setup section handles missing credentials, but polling placeholders like {request_id} and {researchId} are never explained, and there is no poll interval, completion criteria, or error-recovery guidance. Sequence present, checkpoints missing — anchor 3. | 3 / 5 |
Progressive Disclosure | No bundle files exist; six full API catalogs (~390 lines, including complete engine lists and ad-library variants) are inlined in a single SKILL.md with per-API section headers. Structure is present, but content that clearly belongs in per-API reference files is inline, matching anchor 3 rather than anchor 2's 'no section headers'. | 3 / 5 |
Total | 12 / 20 Passed |