Content
76%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 tight, expert-level instruction skill that wastes no tokens and gives concrete, named-API guidance for detection and remediation. The two gaps are the absence of a verification step after batch fix application and the reliance on an out-of-bundle rules file for essential detail. Both are addressable with a small validation step and an inline canonical example.
Suggestions
Add a verification step after applying fixes, e.g. "5. Run typecheck/lint (and tests if present) over the changed files; fix any breakages before finishing" — this removes the batch-operation cap on workflow clarity.
Inline a minimal canonical search-params.ts example (one parseAsString + one createSortParams entry) so the key artifact is copy-paste ready without leaving the skill.
Either bundle the decision table/debounce pattern as a references/ file inside the skill, or note explicitly what to do if .claude/rules/sim-url-state.md is absent.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and dense: it assumes competence with nuqs, React Query, Zustand, and Next.js Suspense without explaining any of them, and every parenthetical carries a substantive exception or carve-out rather than padding. Not 4: there is no over-explanation anywhere that could be trimmed. | 5 / 5 |
Actionability | Guidance is concrete and executable in the main: named APIs ("useQueryState/useQueryStates", "createSortParams + useUrlSort", "useDebouncedSearchSetter"), per-anti-pattern replace-with mappings, and specific fix wiring in step 4. As an instruction-only skill code absence is not penalized, but a minimal search-params.ts example inline would remove the remaining gap, keeping it below the 5 anchor. | 4 / 5 |
Workflow Clarity | Four clearly sequenced steps (read rules → analyze scope → classify via decision table → apply/propose fixes), but the default mode applies fixes across a whole scope (a batch edit operation) with no verification checkpoint such as typecheck/lint/test after applying. Per the scoring notes, a batch operation without validation is capped at 3. Not 2: the sequence itself is complete and well-defined, not gappy. | 3 / 5 |
Progressive Disclosure | Well-organized sections (Context, References, Anti-patterns, Steps) with one-level-deep, clearly signaled references that each describe their contents. It falls short of 5 because the source-of-truth detail (decision table, debounce pattern, sort convention) is deferred to `.claude/rules/sim-url-state.md`, which is not part of the skill bundle and cannot be navigated from the skill itself. | 4 / 5 |
Total | 16 / 20 Passed |