Conciseness

The body does not explain concepts Claude already knows — nearly all 317 lines are novel operational policy — but the same rules are restated multiple times: protected-content gates appear in the Editing contract, the marks-pass paragraph, and the Output format residuals section, and editing-pass accounting is covered in "Iterate to convergence," "Automatic marks pass," and again across ~10 overlapping Output-format paragraphs. This matches "mostly efficient but includes some unnecessary explanation or could be tightened." Not a 2 because the verbosity is redundant policy rather than padded filler or pedagogy, and not a 4 because the cross-section repetition is substantial enough that consolidating would meaningfully shrink the file.

Actionability

Concrete, executable commands are given with exact invocations and exit codes ("node scripts/normalize-quotes.js <rewritten-prose> --reference <original> --write", "node scripts/check-style.js <file> --config <path>" with "exit 0 clean / 1 hard violation / 2 tool error", "node detector/validate.js --residual-policy warn <original> <rewritten>") plus explicit natural-language invocation examples. Not a 5 because the --style config schema is deferred to "examples/README.md", which does not exist in the bundle, and the <rewritten-prose> scratch-file workflow is described abstractly without a worked example — minor gaps that keep it from fully covering the common cases copy-paste ready.

Workflow Clarity

Each mode has an explicit numbered step list (rewrite: Audit/Rewrite/Summarize; detect: Audit/Assess; edit: Read/Edit-in-place/Verify), and validation is everywhere: re-read the file after editing, a four-item Verification checklist (Editing passes, Checks, Residuals, Stop reason), compare-final-with-source accounting, repair-within-remaining-budget feedback loops, and explicit failure reporting when a check cannot run. This matches the anchor-5 pattern of clear sequencing with explicit validation and error-recovery loops; the "Only proceed when validation passes" discipline is fully present.

Progressive Disclosure

The 107KB pattern catalog is correctly split into references/patterns.md, clearly signaled once at the top ("read references/patterns.md in full") and one level deep, and the two invoked scripts are referenced with exact runnable paths that resolve in the bundle (markdown-prose.js is an internal dependency of those scripts, appropriately not surfaced). Not a 5 because the body references paths that do not exist in this bundle — "examples/README.md" ("Schema and rationale: examples/README.md"), "examples/<name>.json", and "detector/validate.js" — and the severity-tier catalog section (P0/P1/P2 pattern lists) partially duplicates material that lives in the reference file. Not a 3 because the references that do exist are clearly signaled, correctly scoped, and easy to navigate.

Specificity

The description lists several concrete actions — "Audit and rewrite content to remove AI writing patterns," "a detect-only mode," "an edit-in-place mode for files," "an optional voice profile (casual / professional / technical / warm / blunt)," "an iterate-to-convergence pass" — matching the anchor for several specific actions with minor gaps (the --style/house-style and --context options are not surfaced). Not a 5 because coverage of the skill's feature set is incomplete rather than comprehensive, and not a 3 because far more than 1-2 concrete actions are named.

Completeness

It explicitly answers both questions with concrete trigger phrases: what ("Audit and rewrite content to remove AI writing patterns ('AI-isms')") and when ("Use this skill when asked to 'remove AI-isms,' 'clean up AI writing,' ...") — mirroring the anchor-5 good example structure exactly. Not a 4 because the 'when' is fully explicit with quoted triggers, not merely present.

Trigger Term Quality

Five natural quoted trigger phrases are given ("remove AI-isms," "clean up AI writing," "edit writing for AI patterns," "audit writing for AI tells," "make this sound less like AI") — good keyword coverage with genuine user phrasing. Not a 5 because common synonyms users actually say are missing (e.g., "humanize this," "make it sound human," "de-slop," "de-AI my text"), which anchor 5's comprehensive-synonym standard expects.

Distinctiveness Conflict Risk

It occupies a clear niche (AI-writing-pattern removal) with distinct, unmistakable triggers like "make this sound less like AI" and "audit writing for AI tells," which no generic editing or proofreading skill would claim. Minimal conflict risk; not a 4 because the quoted triggers leave essentially no overlap ambiguity with related writing skills.