Content
50%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A broad, useful API catalog that is undermined by malformed JSON in many curl examples, heavily repeated auth boilerplate, a monolithic single-file structure with no bundled references, and async workflows lacking completion/validation guidance. It is organized by section but scores at the mid-level across all content dimensions.
Suggestions
Fix the malformed JSON payloads: several -d arguments close the object early (e.g. Tavily /search, Exa /search and /findSimilar, Linkup /search) then append bare keys and a stray '}'. Wrap all request fields inside a single "body": { ... } object so each example is copy-paste executable.
State the shared auth/header block once in Setup and reference it, instead of repeating the full 'Authorization: Bearer $GOOSEWORKS_API_KEY' header block ~40 times, to cut token bloat and improve conciseness.
Split each API into bundled reference files (e.g. references/tavily.md, references/searchapi.md) and keep SKILL.md as an overview with one-level-deep links; also add an explicit completion/status check (what a finished response looks like) to the async deep-research polling steps.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is ~390 lines that repeat the identical auth/header block across roughly 40 curl calls rather than stating it once, which is unnecessary token repetition; it does not explain concepts Claude already knows, so it is above anchor 1, but the repeated boilerplate keeps it from the lean anchor-3 level. | 2 / 3 |
Actionability | Examples are concrete curl commands, but a large fraction have malformed JSON — e.g. Tavily /search, Exa /search, Exa /findSimilar, and Linkup /search close the -d object early with '}' then append bare keys ('"query": ...') and a stray '}', so they are not copy-paste executable; the 'Discover More' block likewise mixes curl with trailing 'api show exa' text. | 2 / 3 |
Workflow Clarity | Async deep-research flows are sequenced as 'Step 1: start / Step 2: poll', but the polling calls use a literal '{request_id}' with no substitution or completion-status guidance, and there are no validation checkpoints; the Tips section gives API-selection guidance but no verify/feedback loop, so it sits at anchor 2 rather than 3. | 2 / 3 |
Progressive Disclosure | There are no bundle files (references/scripts/assets absent) and the entire per-API reference is inline in one ~390-line SKILL.md; section headers (## 1–6, ## Tips, ## Discover More) provide some structure, but content that should be split into separate reference files is monolithic inline, matching anchor 2 rather than the well-signaled multi-file anchor-3 pattern. | 2 / 3 |
Total | 8 / 12 Passed |