Conciseness

The body is lean, directive prose that assumes competence throughout — it never explains what a hypothesis loop, worktree, or todolist is, and lines like 'Rigor is gates and artifacts, not "try harder"' compress policy into single sentences. Matches 'lean and efficient; every token earns its place'; the only candidate for trimming is the intro sentence restating the description.

Actionability

Phases give concrete directives ('Open a todolist whose first item is...', 'State the hypothesis, make the smallest change, measure against the predicate', verdicts 'VERIFIED, NOT VERIFIED, or INCONCLUSIVE'), so guidance is mostly executable. Not 5 because there are no worked examples (of a falsifiable predicate, a log row, or a framing) and sub-steps delegate to external skills without local detail.

Workflow Clarity

A clear Start-then-Phase-A-through-E sequence with explicit validation ('verifying each unit before starting the next', 'Check the whole against the Phase A predicate on the real product') and feedback loops ('keep it if it advanced, revert it if it didn't'; harden the gate when a worker games it). Matches the top anchor; no destructive/batch cap applies.

Progressive Disclosure

The body is under 50 lines with well-organized sections (Start, Phases A-E) and no content that belongs in a separate file; the single file reference (../../docs/harness/claude-code.md) is one level deep and clearly signaled. Under the simple-skill guideline, well-organized sections alone merit the top score; no bundle files exist to restructure.

Specificity

Names the domain ('auditable playbook' for 'a large migration, an ambitious multi-part change') and lists several concrete actions ('Design an auditable playbook', 'runs a hypothesis loop', 'logs decisions'), matching the 'several specific actions; minor gaps' anchor. Not 5 because 'Scales rigor to the task' reads as a property rather than a tangible action; not 3 because multiple distinct actions are enumerated.

Completeness

Explicitly answers 'what' ('Design an auditable playbook... runs a hypothesis loop, and logs decisions') and 'when' with a concrete 'Use for /figure-it-out or \'figure it out\'' clause plus scenario triggers. Matches the anchor for both what and when with concrete trigger phrases; the missing-'Use when' cap does not apply.

Trigger Term Quality

Triggers include natural phrases ('figure it out', 'large migration', 'ambitious multi-part change'), but common variations a user would actually say are missing. Matches 'some relevant keywords but missing common variations or synonyms'; not 4 because coverage of natural phrasings (e.g., 'no clear plan', 'open-ended task', 'big refactor') is thin.

Distinctiveness Conflict Risk

The carve-out 'when no narrower one fits' plus the distinctive 'figure it out' trigger gives a mostly distinct niche with minor overlap risk against planning or migration-specific skills. Not 5 because 'large migration' and 'work a human reviews after stepping away' could still collide with related skills.