Content
75%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 dense, well-structured patterns reference: lean good/bad code pairs, decision trees and matrices, and mostly executable examples covering the common React 18/19 cases. The main improvements are moving long-form examples and the full forms/RSC walkthroughs into reference files, completing the fragmentary composition recipes, and trimming concept-restating commentary.
Suggestions
Move the Examples section and the full form-action/RSC walkthroughs into a references/ file (e.g. references/examples.md) and keep one short representative snippet inline, reducing SKILL.md toward an overview.
Make the composition recipes concrete: give Tabs and compound-component examples minimal working definitions (or note explicitly that they are pattern shapes, not runnable code).
Trim concept-restating commentary ("React has no inheritance model for components", the Routing section) to one actionable line each.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean — good/bad code pairs with one-line rationales, decision trees and matrices instead of prose, no library tutorials or installation padding. Minor over-explanation remains: "React has no inheritance model for components", the Routing section's "follow the router's documentation" filler, and some Core Principles commentary could be trimmed. | 4 / 5 |
Actionability | Mostly executable guidance: near-complete code for useDebounce/SearchBox, the useActionState form, the RSC page, and useOptimistic, plus concrete library names and thresholds ("~50 visible items", "@tanstack/react-virtual"). The composition recipes (Tabs, Layout, DataLoader) are fragments referencing undefined components — pattern-shape illustrations rather than copy-paste-ready code — which keeps it below the anchor-5 bar. | 4 / 5 |
Workflow Clarity | A clear "When to Activate" entry list plus sequenced decision structures (state-location decision tree, data-fetching matrix) give an unambiguous decision path, and no validation checkpoints are needed since the skill involves no destructive or batch operations. It falls short of anchor 5 because the decision paths lack any feedback/checkpoint structure (e.g. how to verify a memoization or boundary placement actually helped). | 4 / 5 |
Progressive Disclosure | Well-sectioned overview with clearly signaled one-level links (rules/react/, sibling skills, agents, commands) and consistent cross-linking. No references/ bundle exists, so all example code and the full RSC/forms guidance are inlined in a ~340-line SKILL.md, and the most-referenced ruleset ("../../rules/react/hooks.md") lives outside the bundle — minor organization gaps versus the anchor-5 clear-split structure. | 4 / 5 |
Total | 16 / 20 Passed |