The body is lean and does not over-explain concepts Claude knows, but the "Common Reveals" and "Anti-Patterns" sections restate findings (inert knobs, singleton wrappers, ceremony) already demonstrated in the worked examples, so it could be tightened. Not a 3 because not every token earns its place; not a 1 because nothing is padded with basic-concept explanation.

Provides concrete audit questions, a numbered 4-step asymmetric-wins checklist, and three ordered reductions with literal fill-in sentences ("This function ______."), fully illustrated by worked examples — actionable instruction rather than vague direction.

Each application has an explicit ordered sequence (sentence → surface audit → asymmetric wins; strip-docs → default-config → caller-need) with explicit stop/finding checkpoints (drift = finding, can't finish = not coherent, caller-need mismatch = verdict). No destructive/batch context, so the cap does not apply.

Self-contained skill with clear, well-organized sections and one-level-deep, clearly signaled links to sibling skills (post-implementation-review, greenfield-clean-breaks, asymmetric-wins); no nested references and no bundle files required.

States the concrete core move ("Force one concrete sentence") and names multiple specific findings ("orphaned surfaces, duplicate verbs, inert abstractions"), matching the 'lists multiple specific concrete actions' anchor.

Explicitly answers both what ("Force one concrete sentence to find…") and when via an explicit "Use for…" clause, plus a negative trigger for when not to use it.

Quotes verbatim natural phrases users actually say — "what does X actually do", "in one sentence", "too many options" — giving good coverage of natural trigger terms.

The one-sentence-reduction niche is distinctive, and the negative trigger ("For a plain code-comprehension question, answer directly instead") reduces the risk of firing for the wrong skill.