Conciseness

The rule index and code examples are tight, but there is unnecessary padding: the 'Ecosystem Coverage' section explains what zustand/leva/postprocessing are (concepts Claude already knows), and the intro line repeats the category count. Not score 2 because the bulk of the body is a lean one-line-per-rule index; not score 4 because the Ecosystem Coverage section and category-table/heading redundancy could be trimmed.

Actionability

The 'Critical Patterns' section gives complete, executable JSX examples (useFrame ref mutation, zustand selectors, useGLTF with preload, Suspense, Perf, visibility toggling) with BAD/GOOD contrast. Not score 5 because the remaining ~70 rules are one-line labels with no code or specifics inline, and the actual detail lives in files that are referenced but not present in the bundle.

Workflow Clarity

As a reference skill its usage is unambiguous: a priority table routes the reader to prefixed rule IDs, and 'How to Use' shows example rule-file paths. Not score 5 because the 'How to Use' section only lists three example paths without stating how to locate a rule for a given task, and the compiled-document path ('../R3F_BEST_PRACTICES.md') introduces ambiguity about where the full content actually lives.

Progressive Disclosure

The body references 'rules/perf-never-set-state-in-useframe.md' and a compiled '../R3F_BEST_PRACTICES.md', but the bundle contains no rules/ directory (nor references/, scripts/, or assets/) — the referenced detail files do not exist, so the progressive-disclosure structure is broken in practice, and the full document path points outside the skill. Not score 3 because the section structure itself is good but the referenced files are simply absent, which is worse than 'not clearly signaled'.

Specificity

The description names the domain ('React Three Fiber (R3F) and Poimandres ecosystem best practices') and lists actions 'writing, reviewing, or optimizing R3F code', but the actions are generic verbs rather than concrete capabilities. It does not score 4 because it never enumerates what the skill actually provides (e.g., performance rules, useFrame patterns, Drei helper usage); it does not score 2 because the domain and several real actions are named.

Completeness

The 'when' is explicit and concrete ('Use when writing, reviewing, or optimizing R3F code. Triggers on tasks involving...'), and the 'what' is present ('R3F and Poimandres ecosystem best practices'). It does not score 5 because the 'what' is a category label rather than concrete statements of what the skill does; it clearly exceeds anchor 3 since a full explicit 'Use when...' clause with trigger packages is present.

Trigger Term Quality

Covers the natural terms users would say: 'React Three Fiber', the synonym 'R3F', and the package names '@react-three/fiber', '@react-three/drei', '@react-three/postprocessing', '@react-three/rapier', 'zustand', and 'leva'. Comprehensive synonym coverage; nothing significant a user would plausibly type is missing.

Distinctiveness Conflict Risk

The R3F/Poimandres niche is highly specific and the triggers are distinct package names unlikely to fire for other skills. Minimal overlap risk — even the closest neighbors (plain Three.js or general React skills) are not claimed.