Content
43%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 well-organized but monolithic API dump: it documents every endpoint with parameters and curl examples inline, yet ships two malformed example commands, no async workflow sequencing (polling/validation for crawls and batches), and no progressive disclosure into reference files. It is serviceable for lookup but fails as an operational guide.
Suggestions
Fix the malformed curl examples for Start Crawl and Start Batch — move the body fields inside the -d JSON payload so the commands are executable as written.
Add explicit async workflows with validation checkpoints, e.g. 'Start crawl → poll GET /v1/crawls/{id} until status=completed → fetch /v1/crawls/{id}/pages → retrieve content by retrieve_id', and equivalent for batches.
Move the bulk per-endpoint parameter reference into a references/ file (or lean on the Discover More details endpoint) and keep SKILL.md to setup, the main workflows, and one example per capability.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly dense parameter lists and curl commands with little concept-explanation padding, but it includes duplication and filler: "Powerful web scraping, crawling, and AI-powered content extraction" restates the description, the Capabilities section repeats the twelve Usage headings verbatim, and long parameter descriptions (e.g., the full default CSS-selector list) could be tightened. This sits at anchor 3 ('mostly efficient but some unnecessary explanation or could be tightened') — above 2 (no heavily padded explanatory prose) but short of 4 due to the redundant capabilities list and boilerplate. | 3 / 5 |
Actionability | Most endpoints have concrete curl commands, but two are syntactically broken: the Start Crawl and Start Batch examples terminate the -d argument at '"path":"/v1/crawls"}' and leave the body fields ("start_url", "max_pages", "items") as orphaned lines outside the quoted JSON, so they fail if copy-pasted. Several endpoints (Batch Items, Crawl Info/Pages, Batch Info, Get Answer, Get Scrape) also show only a '{batch_id}'-style placeholder with no substitution or body/polling detail. This matches anchor 3 ('some concrete guidance but incomplete... missing key details') rather than 4's 'concrete code with minor gaps', since the broken examples are a correctness gap, not just a missing nicety. | 3 / 5 |
Workflow Clarity | The core workflows (start crawl → poll crawl info → fetch pages → retrieve content; start batch → batch info → items → retrieve) are multi-step asynchronous processes, but no sequence is ever laid out — the flow must be inferred from scattered endpoint descriptions — and there is no guidance on polling, checking completion status, or verifying results. Per the guidelines, batch operations without validation steps cap workflow clarity at 3, and this is below that cap: it fits anchor 2 ('rough sequence present but many gaps; validation absent') rather than 3 ('steps listed'), because steps are not listed as a workflow at all. | 2 / 5 |
Progressive Disclosure | The entire ~200-line API reference is inlined in SKILL.md with no references/ bundle and no split into separate files, which matches anchor 3's example ('[200 lines of API reference that could be in a separate file]') — there is clear per-endpoint structure, and the 'Discover More' section offers a live search/details endpoint for deeper info, but the bulk parameter reference clearly belongs in a separate file. Not 2 because the content is well-sectioned and navigable, not a structureless wall; not 4 because no reference files exist and the one-level-deep pattern is absent. | 3 / 5 |
Total | 11 / 20 Passed |