Lean and code-dense — it assumes Claude's competence, never explaining what refactoring or types are, with tight prose and per-section code. Every section earns its place; not score 2, which would require unnecessary explanation it largely avoids.

Concrete executable TypeScript/Svelte snippets, a caller-count decision table, a grep-based straggler sweep checklist with specific search targets, and real commit-message examples. Patterns are copy-paste-ready; not score 2, which would be pseudocode or incomplete.

Clear sequences with checkpoints: 'edit → diagnostics → test → commit → next edit', a numbered post-refactor straggler sweep, and a Go-to-Definition verification step. Commits are test-validated between each, and the sweep is a separate labeled commit, satisfying the destructive/batch feedback-loop requirement.

Single-file skill with no bundle, but well-organized into clearly headed sections; per the simple-skills scoring note this can score 3 without external references. Sibling-skill links are one-level navigation aids, not nested references.

Lists multiple specific concrete actions — 'count callers exactly', 'decide inline vs keep', 'collapse duplicate switch branches', 'route raw access through a single typed boundary', 'surgical one-change-per-commit', 'post-refactor straggler sweep for stale JSDoc and dead exports'. Matches the score-3 anchor listing several concrete actions; not score 2 which only names a domain and some actions.

Explicitly answers both what (the enumerated refactoring mechanics) and when ('Use when actually editing code to clean it up, when deciding whether a 1-caller helper earns its keep, or when planning a sequence of small refactor commits'). Has an explicit 'Use when' clause as the score-3 anchor requires.

Covers natural terms a user would say — 'editing code to clean it up', '1-caller helper earns its keep', 'sequence of small refactor commits', 'refactor'. Good coverage of common variations rather than jargon-only; not score 2 which misses common variations.

Clear niche (per-change refactoring mechanics) with explicit de-conflict routing — 'For the smell catalog use code-audit; for session-long simplification use collapse-pass'. Distinct triggers and low conflict risk match the score-3 anchor.