Content
92%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 well-structured router skill: executable commands dominate, risky operations (linking, mutations) are gated by explicit validation and error-recovery guidance, and the decision tree provides excellent one-level-deep navigation to a complete reference bundle. The only weakness is mild redundancy between the linking section and the Anti-Patterns list, which could be consolidated to trim tokens.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and operational with no explanations of concepts Claude already knows, but there is minor redundancy: linking behaviors (project.json vs repo.json, auto-linking, whoami limits, project inspect verification) are explained in "Critical: Project Linking" and then restated as Anti-Patterns bullets, and the repo-resolution bullets could be tightened. This fits the 'efficient with minor instances that could be trimmed' anchor rather than 'every token earns its place'. | 4 / 5 |
Actionability | Guidance is copy-paste ready throughout: a complete quick-start command block, `vercel project inspect --non-interactive` as a concrete verification command, `vercel deploy --force` without `--with-cache` for no-cache rebuilds, `VERCEL_TOKEN` instead of `--token`, and `vercel curl` for preview access. Specific commands cover the common cases, matching the fully-executable anchor. | 5 / 5 |
Workflow Clarity | Multi-step workflows carry explicit validation checkpoints and feedback loops: before any consequential read or mutation the reader is told to run `vercel project inspect --non-interactive` and confirm owner/project, to stop on `link_required` or target mismatch rather than auto-linking, and the JSON error protocol (`status`, `reason`, `hint`, `next`) provides an explicit error-recovery path with cautions about mutation follow-ups. This matches the anchor for clear sequencing with explicit validation and error-recovery loops. | 5 / 5 |
Progressive Disclosure | The body is a clean overview: an intro with output-parsing rules, a linking gate, a quick start, a decision tree routing each task area to one of 26 real one-level-deep reference files (all verified present in references/), and anti-patterns. References are clearly signaled with topic descriptions, and the few cross-links between reference files are sibling pointers, not nested chains, so navigation stays one level deep. | 5 / 5 |
Total | 19 / 20 Passed |