Content
86%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.
A high-quality, expert-level skill body: nearly every sentence carries a concrete invariant extracted from real integrations, endpoints and contracts are exact, and category-specific detail is correctly delegated to verified one-level-deep references. The only weaknesses are minor token duplication (the contact/review offer appears twice) and the absence of an explicit validate-fix-retry loop in the working workflow.
Suggestions
State the integrations@apify.com contact and review offer once (either the opening callout or the closing paragraph) instead of twice, to trim duplicated content.
Break the long Web Fetch paragraph into a short bullet list of its three gotchas (always send formats, check fetch.httpStatusCode, handle both error shapes) and rely on the linked full contract for detail.
Add an explicit feedback loop to the Working workflow, e.g. 'If the definition-of-done checklist fails, fix the gaps and re-verify before publishing', to close the validation cycle.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and rule-driven with no explanations of concepts Claude already knows (the glossary covers genuinely non-obvious Apify vocabulary like "Actor - Not an AI agent" and "may be still called Standby"), but the integrations@apify.com contact/review offer is stated twice (opening callout and closing paragraph) and the Web Fetch section is a single overlong paragraph - minor instances of padding that could be trimmed. | 4 / 5 |
Actionability | Guidance is fully executable: exact endpoints ("POST /v2/actors/{actorId}/runs"), exact header values ("x-apify-integration-ai-tool: true"), a copy-paste Web Fetch contract ("POST https://web-fetch.apify.actor/ with {\"url\": \"...\", \"formats\": [\"markdown\"]}"), and canonical cost parameters with explicit placement rules ("Send them as options / query parameters, never as Actor input"). | 5 / 5 |
Workflow Clarity | The "Working workflow" gives a clearly sequenced 6-step process with a required Step 0 and a terminal "Verify against the definition-of-done checklist" checkpoint, plus validation via two test modes (mocked and live E2E). It falls short of anchor 5 because there is no explicit fix-and-re-verify feedback loop if the checklist or tests fail. | 4 / 5 |
Progressive Disclosure | The body is a cross-cutting overview that routes to exactly one of four one-level-deep reference files via an unambiguous product-shape table ("Read exactly one reference file"); all four files exist in references/ and point back to SKILL.md rather than nesting further, so navigation is easy and content is appropriately split. | 5 / 5 |
Total | 18 / 20 Passed |