Content
38%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 self-contained API catalog with real curl commands and Tavily-specific parameter detail, but it inlines ~150 lines of reference material that belongs in separate files, repeats parameter descriptions verbatim, and ships two malformed curl examples. Restructuring into an overview plus reference files and fixing the broken examples would address the weakest dimensions.
Suggestions
Move the per-endpoint parameter reference into a references/ file (e.g. references/endpoints.md) and keep SKILL.md as a concise overview with one working example per capability and clearly signaled one-level-deep links.
Fix the malformed curl examples for Tavily Search and Tavily Crawl so all parameters are inside the -d JSON body (e.g. -d '{"api":"tavily","path":"/search","body":{"query":"...","search_depth":"advanced","include_answer":true}}').
Deduplicate the chunks_per_source and credit-cost text repeated across Search/Extract/Crawl, and spell out the research workflow as an explicit sequence: create task → save request_id → poll Get Research Task Status until complete.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~150-line parameter dump is noticeably verbose: the chunks_per_source description is repeated nearly verbatim in the Search, Extract, and Crawl sections, and credit-cost prose pads multiple parameters. This goes beyond the minor trimming of anchor 3 into repeated, padded sections. | 2 / 5 |
Actionability | Setup and the Extract/Map/Research curl examples are concrete and executable, but the Tavily Search example (lines 62-66) and Tavily Crawl example (lines 164-167) have malformed JSON — parameters appear outside the -d request body string — leaving key details broken. This is more than the minor gaps of anchor 4. | 3 / 5 |
Workflow Clarity | Capabilities are catalogued per-endpoint but never sequenced: the create-research-task → poll-status-by-request_id flow is implied rather than stated, and there is no error-handling or verification guidance for batch crawl/map operations. Anchor 3 (steps/sequence present but checkpoints missing) fits; not 4 since no workflow is actually laid out. | 3 / 5 |
Progressive Disclosure | No bundle files exist and the entire API parameter reference is inlined in SKILL.md, which clearly belongs in separate reference files; sections exist but the bulk reference is not split or offloaded, matching anchor 2. | 2 / 5 |
Total | 10 / 20 Passed |