Conciseness

Every section carries library-specific, non-obvious knowledge — in-place mutation semantics ('Transforms mutate the document in place'), silent failure of invalid tags ('an invalid #Tag matches nothing, silently'), term-level match syntax, and methods that throw ('.money().currency()… do not exist'). There is no general NLP/JavaScript explanation and no padding; each token earns its place. Not 4 because there are no identifiable over-explanations to trim — even the intro paragraph conveys the library-specific document/View model, not concepts Claude already knows.

Actionability

Fully executable, copy-paste-ready code with expected outputs throughout ('nlp(\'five hundred\').numbers().toNumber().text() // \'500\''), including correct-vs-incorrect contrasts for the mutation rule, exact replacements for nonexistent methods ('Use .money().json() / .fractions().get() instead'), and concrete debugging commands. Matches the anchor-5 pattern of runnable examples covering the common cases; not 4 because no key case is left pseudocode or incomplete.

Workflow Clarity

A single-purpose skill whose flow is unambiguous: internalize the five rules → apply verified patterns → heed sharp edges → follow the explicit error-recovery loop ('doc.debug() // prints how every word was actually tagged — start here', then json(), verbose, then a three-cause checklist for empty matches). The debugging section is a genuine validate→diagnose→retry feedback loop; not 4 because checkpoints for the risky paths (wrong tags, empty matches, throwing methods) are explicit rather than implicit.

Progressive Disclosure

Good structure: a lean cheat-sheet body with a well-signaled 'Going deeper' section pointing one level deep to version-matched package docs, each described by purpose ('match-syntax.md — every match operator', 'tag-definitions.md — the complete, valid tagset'). Not 5 because the inline pointer '(see references for the rest)' does not resolve to a real bundle file — no references/ directory ships with the skill — leaving one navigation target ambiguous; not 3 because all other references are clearly signaled and the content split is otherwise appropriate.

Specificity

Base score is 5: the description lists multiple concrete actions — 'extracting entities/dates/numbers, matching text patterns, changing verb tense, pluralizing, redacting/anonymizing text, building a simple chatbot intent matcher' — with comprehensive coverage of the library's capability surface. However, the second-person phrasing 'Use this whenever you are writing or editing JS/TS' violates the third-person voice rule, triggering the −1 penalty; it is not a 3 because the action list itself is comprehensive and concrete, far beyond '1-2 concrete actions'.

Completeness

Explicitly answers both questions: what ('Write correct code with the compromise JavaScript NLP library…') and when, with concrete trigger phrases ('Use this whenever you are writing or editing JS/TS that imports compromise, calls nlp(...)… and also whenever doing rule-based natural-language tasks in JavaScript'). Matches the anchor-5 example's structure exactly; not 4 because the 'when' clause is not merely present but enumerates specific triggering conditions.

Trigger Term Quality

Comprehensive natural-term coverage: the package name and import form ('the npm package compromise, imported as nlp'), API surface terms ('nlp(...)', '.match()', '.people()', '.verbs()'), and task phrases users would actually say ('extracting entities', 'changing verb tense', 'redacting/anonymizing text', 'chatbot intent matcher'). It sits at the anchor-5 example's level (PDF files, PDFs, forms, .pdf) rather than anchor 4, which leaves natural synonyms missing — none are missing here.

Distinctiveness Conflict Risk

The niche is distinct — a specific npm library with signature calls ('imports compromise, calls nlp(...)') and named methods — but the catch-all clause 'whenever doing rule-based natural-language tasks in JavaScript… where compromise is or could be the tool' broadens triggers to any rule-based NLP task in JS, creating minor overlap risk with general text-processing/NLP skills. Not 5 because that broadened trigger could fire for tasks where a different tool is intended; not 3 because the primary triggers are unmistakably library-specific.