Conciseness

The body is a dense rule list with no filler and no explanation of concepts Claude already knows: each rule gives a pattern, quoted example phrases, and a fix ("'In order to' becomes 'To'", "'utilize' becomes 'use'"). Every token carries a detection cue or a fix, matching the lean-and-efficient anchor.

Actionability

Every rule is executable editing guidance with concrete before/after examples: quoted tells ("'serves as', 'stands as', 'boasts'") and specific replacements ("'queries are validated' becomes 'the compiler validates queries'"). The fixes are stated for each pattern, covering the common cases directly.

Workflow Clarity

The Process section is a clear three-step sequence (Scan for the patterns, Rewrite while preserving meaning and tone, then "Self-audit: 'What makes this obviously AI generated?' Fix remaining tells") with an explicit validation checkpoint and feedback loop. The categorized rule list functions as a checklist. This is not a destructive or batch operation, so no cap applies.

Progressive Disclosure

A single well-sectioned file (Content, Language, Style, Communication artifacts, Filler, Jargon, Plain speech) with no bundle files to reference. All 28 rules live inline at roughly 60 lines of body, just past the point where the under-50-line simple-skill exception would grant a 5, and the rule catalog is the kind of detailed material a longer skill would split into a reference file. Good structure with a minor organization consideration keeps it at anchor 4.

Specificity

"Cut AI tells from any writing" names one concrete action with a clear object, but lists no capability breadth (no mention of which patterns, vocabulary, or structure it fixes). This matches the anchor for naming the domain plus 1-2 concrete actions without comprehensive coverage; a 4 would require several listed actions, and a 2 would require a generic action like "Processes files".

Completeness

The "what" is clear ("Cut AI tells from any writing"), but the only when-guidance is "Must always apply", a blanket applicability claim rather than a situational trigger. It is present but degenerate, so this sits at the clear-what/weak-when anchor rather than anchor 4, which expects a real conditional trigger like "Use when working with PDF files".

Trigger Term Quality

"AI tells" and "writing" are relevant keywords, but no synonyms or natural user phrasings appear (nothing like "sound less AI", "AI slop", "chatbot tone", "humanize"). This matches "some relevant keywords but missing common variations or synonyms"; a 4 would need several natural trigger terms.

Distinctiveness Conflict Risk

"Cut AI tells" names a distinct niche, but "from any writing" plus "Must always apply" claims applicability to all text, creating overlap risk with any writing or editing skill. This matches "somewhat specific but could still overlap with similar skills"; the task focus keeps it above the very-broad anchors 1-2.