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 well-sectioned API reference with concrete curl examples, but two primary examples contain broken JSON, the async poll workflow is never made explicit, and the full endpoint reference is monolithic with no bundle files. It lands at the midpoint across dimensions.
Suggestions
Fix the malformed JSON in the Start SmartScraper and Start SmartCrawler examples so the body fields sit inside the `-d` object, matching the correct `"body":{...}` form used by SearchScraper/Scrape/Sitemap/Markdownify.
Add an explicit async workflow with a validation checkpoint: start the request, then poll the corresponding 'Get ... Status (free)' endpoint until it returns a terminal state, and surface what to do on failure.
Fill in descriptions for the Start SmartCrawler parameters (url, prompt, extraction_mode, depth, max_pages, etc.) and move the full per-endpoint parameter reference into a separate reference file linked from SKILL.md.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body avoids explaining concepts Claude already knows and is organized per-endpoint, but the curl header boilerplate and identical stealth/mock/headers explanations are repeated across six endpoints, so it is mostly efficient with tightening opportunities. | 3 / 5 |
Actionability | It provides concrete curl commands per endpoint, but the primary SmartScraper and SmartCrawler examples are malformed JSON (body fields sit outside the `-d` object) and SmartCrawler parameters lack descriptions, leaving key examples non-executable. | 3 / 5 |
Workflow Clarity | The async start-then-poll pattern is only implicit, with no explicit poll-until-complete steps, status check thresholds, or error-recovery guidance; because crawling is a batch operation, the missing feedback loop caps this at 3. | 3 / 5 |
Progressive Disclosure | Sections are clearly headed and the 'Discover More' block points to the live search/details API, but all eleven endpoints are fully detailed inline in one ~230-line file with no separate reference files, so content that could be split stays inline. | 3 / 5 |
Total | 12 / 20 Passed |