Content
72%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 tight, project-specific enforcement checklist with concrete file paths, expressions, and helper names, and clean one-level-deep reference delegation. Its main weakness is the fix=true workflow, which applies batch edits with no validation or feedback loop.
Suggestions
Add a validation checkpoint to the fix=true path, e.g. "After applying fixes, run the typecheck/tests and review the diff; fix and re-verify on failure."
Include one before/after violation example (e.g., a component copying query data into useState) to anchor what a finding and its fix look like.
Split the long key-factory bullet into two or three shorter bullets so each rule is independently scannable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes Claude's competence — it never explains what React Query is and states only project-specific constraints ("Server data must never be duplicated into useState or Zustand outside of mutation callbacks"). It is not 5 because a few passages could be trimmed: the long run-on key-factory bullet and the parenthetical "(fires on success and error)" explain semantics Claude already knows. | 4 / 5 |
Actionability | Guidance is highly concrete: exact paths (".claude/rules/sim-queries.md", "hooks/queries/utils/prefetch-query-on-intent.ts"), a copy-paste-ready expression ("Boolean(id) && (options?.enabled ?? true)"), and named helpers ("prefetchQueryOnIntent(queryClient, sharedQueryOptions)"). It is not 5 because there are no before/after violation examples or an expected output format for the audit findings — minor gaps for an instruction-only skill. | 4 / 5 |
Workflow Clarity | The steps are clearly sequenced ("1. Read the references... 2. Analyze the specified scope... 3. If fix=true, apply the fixes") but the fix=true path is a batch edit operation with no validation checkpoint (no typecheck, test run, or diff review before finishing), which caps workflow clarity at 3 per the batch-operation rule. It is not 4 because validation is entirely absent rather than a minor gap. | 3 / 5 |
Progressive Disclosure | The ~47-line body is well organized into Context, References, Rules, and Steps, and delegates detail one level deep to clearly annotated references ("https://tkdodo.eu/blog/effective-react-query-keys — key factory pattern, hierarchical keys, fuzzy invalidation") and to project rule files. No bundle files exist or are needed, so the under-50-line well-organized-sections standard for a 5 applies. It is not 4 because navigation is clean and nothing that belongs in a separate file is inlined. | 5 / 5 |
Total | 16 / 20 Passed |