Content
78%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 highly actionable, well-sectioned API reference with executable examples for every access path (MCP, REST, SDK, CLI) and a useful MCP-vs-CLI-vs-REST decision table. Its weaknesses are the monolithic single-file structure — a large endpoint table and multiple SDK examples that should live in reference files — and minor marketing padding.
Suggestions
Move the 18-row endpoint table and the SDK code examples into references/ files (e.g. references/endpoints.md, references/sdk-examples.md), keeping SKILL.md to connection setup, the decision table, and cross-references.
Delete non-actionable marketing lines such as 'Used in production by Zapier, Composio, Vapi, and Solana'.
Add an explicit verification step to the write-oriented guidance (e.g. after creating an env var, re-list env vars to confirm it appears in the target environments).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense reference material — endpoint tables, executable curl/SDK/CLI snippets, and a decision table — with almost no explanation of concepts Claude already knows. Minor trimmable padding keeps it below 5: 'Used in production by Zapier, Composio, Vapi, and Solana' is marketing fluff, and the intro paragraph ('You are an expert in...') adds little. Not 3 because none of the sections are genuinely unnecessary explanation. | 4 / 5 |
Actionability | Guidance is fully executable and copy-paste ready across every access path: a working curl with auth header, complete @vercel/sdk snippets (list deployments, create env var with target/type fields, list domains, create log drain), a REST endpoint table with methods, and exact 'vercel api' CLI invocations including a --body example. Common cases are covered with specific, runnable code. | 5 / 5 |
Workflow Clarity | The usage patterns give clear, coherent numbered sequences ('List recent deployments → find the failed one / Inspect deployment → get error summary / Query build logs → identify root cause'), each ending in an implicit checkpoint (root cause, verified domain assignment). It stays at 4 rather than 5 because the write-oriented guidance (create env var, create/delete drain) includes no verification step after mutation. | 4 / 5 |
Progressive Disclosure | No bundle files exist (no references/, scripts/, or assets/), and the ~255-line body inlines substantial material that clearly belongs in separate reference files — the 18-row endpoint table, four full SDK examples, and the drains payload schemas (deferred only by a 'see ⤳ skill: observability' pointer, which is one level deep and well signaled). Section headers and cross-references are well organized, but the bulk API reference inline matches the 'could be better organized' anchor rather than an appropriately split structure. | 3 / 5 |
Total | 16 / 20 Passed |