Conciseness

Dense, high-signal content — a match-syntax cheat-sheet table, verified code patterns, and sharp-edge warnings — with no generic concept padding; the brief opening framing ('rule-based… no network, no model, no deps… jQuery-like chained API') is library-specific and earned. It assumes Claude's competence rather than re-explaining basics, matching the lean score-3 anchor.

Actionability

Code is fully executable with expected outputs and inline ✅/❌ annotations — e.g. 'doc.verbs().toPastTense(); doc.text() // I walked' and 'nlp("five hundred").numbers().toNumber().text() // 500'. Copy-paste ready, clearing the score-3 bar over the pseudocode of score 2.

Workflow Clarity

The 'Debugging a wrong result' section gives a clear sequenced feedback loop (doc.debug() → doc.json() → nlp.verbose(true), then check tag validity / sentence boundaries / {root} conjugations). There is no destructive or batch operation requiring validation, so the well-organized sequence scores 3.

Progressive Disclosure

SKILL.md is a well-organized overview (rules, common tasks, sharp edges, debugging) and the 'Going deeper' section cleanly signals one-level-deep pointers to the version-matched package docs (node_modules/compromise/docs/*.md) with no nested references. Navigation is easy and content is appropriately split.

Specificity

Lists multiple concrete actions — 'writing or editing JS/TS that imports compromise, calls nlp(…), or chains methods like .match(), .tag(), .people()…' plus 'extracting entities/dates/numbers… changing verb tense, pluralizing, redacting/anonymizing text, building a simple chatbot intent matcher'. This matches the score-3 anchor of multiple specific concrete actions, not the partial coverage of score 2.

Completeness

Explicitly answers both 'what' ('Write correct code with the compromise JavaScript NLP library') and 'when' ('Use this whenever you are writing or editing JS/TS that imports compromise…'). The explicit 'Use this whenever…' trigger clears the cap that a missing 'Use when' clause would impose at 2.

Trigger Term Quality

Covers natural user phrasing — 'imports compromise', 'calls nlp(…)', '.match()/.verbs()/.nouns()/.numbers()/.normalize()/.replace()', 'extracting entities', 'changing verb tense', 'pluralizing', 'redacting/anonymizing'. Good coverage of terms users would actually say, clearing the score-3 bar rather than the partial score-2.

Distinctiveness Conflict Risk

A clear niche — the compromise library with specific method-chain triggers (nlp(…), .match(), .tag(), .people()) — makes it unlikely to fire for unrelated skills. Distinct enough to avoid the overlap risk of score 2.