Conciseness

Every line is either compressed doctrine ("A repeated correction is an environment bug. Fix the kitchen, not the dish."; "Cost is what you pay to write it plus what you pay to keep it true.") or concrete instruction, with no explanation of concepts Claude already knows — mutation survivors, analyzers, and gh are used without preamble. The brief philosophical lines are load-bearing heuristics rather than padding, matching the "every token earns its place" anchor.

Actionability

The evidence step is copy-paste ready (gh pr list, gh run list, scan.py, run-mutation-smoke.sh, the gh api loop with REPO assigned) and the layer table gives exact paths (.editorconfig, BannedSymbols.txt, src/Zipper.Analyzers/, tests/goldens/) plus existing analyzer IDs ZIP001–ZIP004. Minor gaps keep it below a 5: the core deliverable — writing the analyzer/test/hook — has concrete procedures ("a case in src/Zipper.Analyzers.Tests/ that fails on the bad shape and passes on the fixed shape") but no example code, and PR must be set manually mid-block.

Workflow Clarity

A clear four-step sequence (collect evidence, pick the narrowest layer, prove the constraint fires, land one change) with explicit validation throughout: "Do not fix the individual instances yet", "test: red first, then implement", "make a scratch change that violates the rule, run the gate, paste the failure, then revert". It even handles the edge case where validation cannot exercise itself ("the pre-push gate exits green without running the new operator. Prove it manually"), and the Red flags section acts as a checklist — matching the top anchor with feedback loops and no missing checkpoints.

Progressive Disclosure

Clean single-level structure with well-organized sections ("## When this applies", numbered steps 1–4, "## Red flags") and nothing buried or nested. It scores 4 rather than 5 because at ~117 lines it exceeds the under-50-line simple-skill exception, and repo-specific catalogs (the layer table, the analyzer ID registry, the evidence commands) are inlined in one file rather than split into one-level-deep reference files.

Specificity

The description names the domain and a single action — "to turn the pattern into a permanent constraint instead of another prose rule" — but "permanent constraint" is abstract and names no mechanism (analyzer, banned symbol, hook, mutation operator, test). This matches the anchor for 1–2 concrete actions that are not comprehensive, below anchor 4 which expects several specific actions listed.

Completeness

Both halves are present: the 'when' is explicit and concrete ("Use when the same agent mistake, review finding, mutation survivor, or CI failure appears twice, or when spot-checking recently merged PRs"), and the 'what' states the transformation ("turn the pattern into a permanent constraint instead of another prose rule"). It falls short of anchor 5 only because the 'what' is abstract — no concrete deliverable or mechanism is named.

Trigger Term Quality

Natural trigger phrases are strong and varied — "agent mistake", "review finding", "mutation survivor", "CI failure", "appears twice", "spot-checking recently merged PRs" — phrases a user in this workflow would genuinely say. A few natural variants are missing ("recurring", "regression", "keeps happening"), so it does not reach the comprehensive-synonym coverage of anchor 5.

Distinctiveness Conflict Risk

The description occupies a clear niche — recurrence-driven environment hardening — anchored by distinctive terms like "mutation survivor" and the explicit twice-occurrence condition, which generic review, CI, or testing skills would not claim. Mis-trigger risk is minimal, matching the top anchor.