The body is dense and information-rich without explaining basics Claude already knows; concrete commands like `git diff`/`git status` earn their tokens and the prose assumes competence throughout.

Gives concrete executable guidance - `git diff`, `git status`, `/fable-judge suite <target>` - plus a precise ordered procedure and a concrete catalogue of frauds to hunt, fully actionable for an instruction-style skill.

The default mode is a clearly numbered 1-5 sequence with an explicit validation step ("Re-run every claimed verification yourself") and an evidence-first verdict step; suite mode is similarly sequenced.

Keeps the detailed fraud catalogue and domain adapters external in fable-method's `references/failure-modes.md` and `references/domains/`, with clearly signaled one-level-deep pointers rather than inlining them.

Lists multiple concrete actions - "re-runs the claimed verifications", "diffs what actually changed", "detects weakened tests and false completion claims", and "delivers an evidence-based verdict" - matching the anchor for multiple specific concrete actions.

Clearly states both what it does (adversarial verification of completion claims) and when to use it ("Use after any agent or model claims work is complete") with explicit triggers, satisfying the both-what-and-when anchor.

Provides natural trigger phrases a user would actually say: "/fable-judge", "judge this work", "verify what it did", "did that actually work?", giving good coverage of natural terms.

A clear niche - adversarial verification of finished work - with distinct triggers unlikely to fire for unrelated skills, matching the clear-niche anchor.