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 well-structured, actionable skill body that assumes Claude's competence and gives concrete file/symbol-level guidance. The main defect is the missing reference.md template, which breaks the only external reference and caps progressive disclosure.
Suggestions
Create the referenced reference.md (or place the full template in references/) so the 'Full template: see reference.md' link resolves, or inline the template and drop the reference.
Add an explicit validation feedback loop in Build & verify (e.g. 'if diagnostics report type errors, fix and re-run diagnostics before confirming connected').
Deduplicate the 'connected vs credentialed' guidance between the 'How it works' callout and 'Conventions & gotchas' to tighten conciseness.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Dense and assumes Claude's competence (no explaining of TanStack Query or basic React), but the 'Connection is detected, not asserted' point is restated in Conventions & gotchas, a minor trim opportunity. | 4 / 5 |
Actionability | Highly actionable instruction-only content: exact file paths, full prop contract, a 6-step state machine, specific query keys to invalidate, and named primitives — guidance is concrete enough to execute directly. | 5 / 5 |
Workflow Clarity | Clear sequence (prerequisites → file checklist → state machine → recipe → build & verify) with a validation checkpoint ('check types via Cursor diagnostics'), but the error-recovery feedback loop is implicit rather than explicit. | 4 / 5 |
Progressive Disclosure | Well-organized inline sections with a clearly signaled one-level reference ('Full template: see reference.md'), but reference.md does not exist in any bundle directory, so the reference is a dead link that harms navigation. | 3 / 5 |
Total | 16 / 20 Passed |