Conciseness

Every line is a lean, directive instruction with no padding and no explanation of concepts Claude already knows (e.g., no primer on what foreign keys or migrations are). Matches the anchor 'every token earns its place'. Not a 4 because there are no instances of over-explanation to trim.

Actionability

Concrete, specific guidance throughout (e.g., "check row counts and checksums before and after", "delete only rows tagged with the run ID", "Log every deletion and every failure in cleanup.csv", "A 200 response from DELETE is not proof") — actionable for an instruction-only skill per the scoring notes. Not a 5 because no example queries/commands are given and "(pattern G1)" references an undefined external pattern the reader cannot resolve.

Workflow Clarity

Steps 1–7 are a clear, ordered sequence with an explicit validation checkpoint (step 7: query leftover records by tag) and exit criteria ("Integrity queries return zero violations"), so the destructive-operation cap does not apply. Not a 5 because there is no feedback loop for error recovery — the body never says what to do when violations or leftover records are found.

Progressive Disclosure

The skill is 16 lines with no need for external references (none of references/, scripts/, assets/ exist), and content is organized under clear 'Procedure' and 'Exit criteria' sections — per the guideline, under-50-line skills with well-organized sections can score 5. The dangling "(pattern G1)" mention is a token-level blemish already reflected in actionability, not a structural organization gap.

Specificity

"Verify persistence, constraints, migrations, duplicates, orphans, and cleanup of test data" names six concrete, specific actions with comprehensive coverage, matching the anchor for multiple specific concrete actions. Third-person verb form ("Verify") is correct. Not a 4 because coverage is comprehensive rather than having minor gaps.

Completeness

Explicitly answers both: what ("Verify persistence, constraints, migrations, duplicates, orphans, and cleanup of test data") and when ("Use when testing anything that writes data or after schema changes and deploys") with concrete trigger phrases, exactly matching the anchor. Not a 4 because the 'when' clause is explicit rather than merely adequate.

Trigger Term Quality

Natural terms users would say are present: "test data", "schema changes", "deploys", "duplicates", "orphans", "cleanup". Not a 5 because common variations like "database", "data integrity", "leftover/stale data" are missing; not a 3 because the existing keyword set is broad and natural, not just one or two generic terms.

Distinctiveness Conflict Risk

The test-data-integrity niche (orphans, retention, cleanup ledger, migration checks) is fairly distinct with targeted triggers. Not a 5 because "anything that writes data" is broad and could overlap with general testing or database-verification skills; not a 3 because the specific what-clause keeps overlap risk minor and scoped to closely related skills.