Conciseness

Lean and efficient: every rule earns its place with inline before/after examples ("'In order to' becomes 'To'", "'utilize' and 'leverage' become 'use'") and zero padding or explanation of concepts Claude already knows. It is not 4 because there are no over-explanations that could be trimmed; the one longer rule (14) spends its words on a worked example that directly disambiguates a subtle case.

Actionability

Fully actionable instruction-only guidance: nearly every rule pairs a detection pattern with a concrete fix or literal replacement ("'queries are validated' becomes 'the compiler validates queries'", "break a noun cluster longer than three words"), and rule 27 even supplies a usable self-test ("if the sentence could appear unchanged in another project's docs, cut it"). For an instruction skill this is copy-paste-grade direction with no gaps.

Workflow Clarity

A simple single-task skill with an unambiguous two-part instruction: "Scan for the patterns below, then rewrite. Preserve meaning and match the intended tone." Per the simple-skill allowance, a single unambiguous action can score 5; prose editing is not a destructive or batch operation requiring validation checkpoints, so the cap does not apply. Not 4 because nothing in the sequence is ambiguous or missing.

Progressive Disclosure

A single flat self-contained file with no nested references and well-organized category headers (Content, Language, Style, Communication artifacts, Filler, Jargon, Plain speech), plus stable rule ids for citation by other skills. Anti-drift check against anchor 4: there are no "minor organization gaps" and nothing that belongs in a separate file; the whole rule catalog is the skill's operational core and loads in one screen.

Specificity

Names the domain and action ("Cut AI tells from prose people read") and enumerates several concrete targets: "PR titles and bodies, Slack and Jira posts, READMEs and other human docs". It is not 5 because it offers one core action plus a scope exclusion rather than a comprehensive action list, and not 3 because it goes beyond 1-2 generic actions by naming multiple specific artifact types. Third-person voice, so no voice penalty applies.

Completeness

The "what" is clear, but there is no "Use when..." clause or equivalent explicit trigger guidance; the "when" is only weakly implied by the artifact list. Per judging guidelines, a missing 'Use when...' clause caps completeness at 3. Not 4 because the "when" is not present in any explicit form.

Trigger Term Quality

Good natural keyword coverage of target surfaces ("PR titles and bodies", "Slack and Jira posts", "READMEs", "human docs") that users would plausibly say. It is not 5 because common variations of the core request are missing ("AI slop", "sounds AI-generated", "make it sound human"), and not 3 because coverage is good rather than merely "some relevant keywords".

Distinctiveness Conflict Risk

A distinct niche (removing AI tells from human-facing prose) with an explicit scoping exclusion ("Leave the fixed headings and bold labels of the `pr` skill and of PR templates as they are"). Not 5 because without trigger phrases it could still overlap with general writing/editing skills. Not 3 because the niche is well-differentiated, not just "somewhat specific".