Conciseness

The body is dense payload Claude does not already know (33 patterns with before/after pairs, false-positive guidance), but the never-invent-facts rule is repeated at least five times ('Never invent facts', 'Never invent one to make a sentence sound sourced', 'come from sources or the user, not from the rewrite', the Process step 3 audit question) and the em-dash sample exception is restated three times. This fits the 4 anchor ('minor instances of over-explanation that could be trimmed') rather than 5's 'every token earns its place'.

Actionability

For an instruction-only skill the guidance is fully concrete: all 33 patterns carry verbatim Before/After pairs, section 14 gives an executable final check ('scan it for `—` and `–`. Any hit means the draft isn't done'), and the two audit questions are quoted word-for-word. Specific examples cover the common cases, matching the 5 anchor rather than 4, since no key execution detail is missing.

Workflow Clarity

'Process and Output' sequences draft, audit, final with explicit validation checkpoints (the two named audit questions and the em/en-dash scan) and a revise-to-address feedback loop, and each invocation mode states exactly what to deliver. This matches the 5 anchor ('clear sequence with explicit validation steps; feedback loops'); 4 would require a missing checkpoint, and none are absent.

Progressive Disclosure

Internal structure is well sectioned and the Strunk reference is clearly signaled with size guidance ('~30,000 tokens. Read it only when you are actively writing or editing prose'), but the referenced `elements-of-style.md` does not exist anywhere in the bundle (no references/ directory and no such file in this directory), so the single external reference is a dead end, and the full 33-pattern catalog is inlined in a roughly 450-line body. This fits the 3 anchor ('some structure but could be better organized; content that should be separate is inline') better than 4, whose references must be at least 'mostly clear' with only minor gaps.

Specificity

Concrete actions are enumerated throughout: 'Strip AI-writing tells', 'make prose clearer, stronger, and more concise', 'Detects and fixes AI patterns (inflated symbolism, promotional language, superficial -ing analyses...)', and 'applies Strunk's Elements of Style rules for clarity and concision (active voice, positive form, definite and concrete language, omitting needless words, parallel construction)'. Multiple specific actions with comprehensive coverage matches the 5 anchor; 4 would require minor gaps, but both the pattern list and the style rules are itemized. The description uses third-person/imperative voice, so no person penalty applies.

Completeness

Both questions are answered explicitly: the 'what' ('Strip AI-writing tells AND make prose clearer, stronger, and more concise... Detects and fixes AI patterns... applies Strunk's... rules') and the 'when' via an explicit 'Use when editing or reviewing text... or when writing or editing any prose humans will read' clause with concrete trigger contexts. This matches the 5 anchor ('clearly and explicitly answers both what AND when with concrete trigger phrases') exactly.

Trigger Term Quality

Natural keywords are strong: 'editing', 'reviewing text', 'sound natural and human', plus a long enumeration of prose types ('documentation, commit messages, pull request descriptions, error messages, UI copy, help text, comments, reports, or explanations'). A few natural user phrasings such as 'rewrite' or 'tone' are absent, so this sits between the 4 anchor ('good keyword coverage; a few natural terms missing') and the 5 anchor's fully comprehensive synonym coverage.

Distinctiveness Conflict Risk

The AI-tell-removal niche ('AI-writing tells', 'AI patterns', 'sound natural and human', 'Strunk's Elements of Style') is clearly distinguishable from most skills, but the trigger surface 'any prose humans will read' with a nine-item enumeration is broad enough to overlap with general writing, editing, and documentation skills. This fits the 4 anchor ('mostly distinct; minor overlap risk with closely related skills') rather than 5's minimal conflict risk.