Conciseness

The overwhelming majority of the body is dense, high-value executable test code, but there are minor instances that could be trimmed: the Overview's explanation of EXPIRE/PEXPIRE/EXPIREAT basics, the throwaway claim "Redis is the dominant application-tier cache", and TTL return-code semantics (-1/-2) stated three times (test comments, the 'Parsing results' section, and the Overview). Not anchor 3 because the unnecessary prose is a small fraction of the document; not anchor 5 because the duplication and basic-concept explanation are real trimming targets.

Actionability

Mostly copy-paste-ready pytest/YAML: session-scoped testcontainers fixture, TTL/conditional-expire/eviction/stampede/tenant/pub-sub tests, run commands, and CI service config. Not anchor 5 because two examples depend on undefined app-layer fixtures ("db", "cache", "update_user_via_app") and so are not fully executable as written — minor gaps matching anchor 4.

Workflow Clarity

The install → fixture → write-pattern-tests → run → CI sequence is discernible, and most checkpoints are present (per-test flushdb, range-assertion guidance, CI "--health-cmd 'redis-cli ping'", an anti-patterns checklist). Minor validation gaps keep it below anchor 5: the eviction test mutates "maxmemory"/"maxmemory-policy" with no restore step, and the 'Pub-sub invalidation' section is misplaced under '## Running' after the run command.

Progressive Disclosure

Referenced paths are real, one level deep, and clearly signaled both inline and in a References section, with background appropriately split out of the body. Not anchor 5 because 4 of the 6 bundle reference files (stale-while-revalidate.md, rfc-9111-http-caching-directives.md, stampede-mitigations.md, cross-tier-coherence-and-test-surface.md) are never linked from SKILL.md — including stale-while-revalidate.md, which the description explicitly promises — leaving that content undiscoverable; not anchor 3 because the structure and signaling of what is referenced is good.

Specificity

Enumerates several concrete coverage areas ("TTLs, cache-aside write and invalidate, eviction under memory pressure, pub/sub invalidation across nodes and tenant key namespaces, plus stampede and stale-while-revalidate patterns"), but the items are topic nouns rather than verb-led actions and the skill's own action is only implied by "its behavior needs tests" — a minor gap that fits anchor 4 better than the comprehensive verb-action anchor 5.

Completeness

The "when" is explicit and concrete ("Use when Redis is the app's cache and its behavior needs tests, or when choosing a cache invalidation strategy"), and the "what" (testing the enumerated behaviors) is present but embedded in the when-clause rather than stated as a leading capability sentence. Both are present with the 'what' less explicit, matching anchor 4; it does not reach anchor 5's 'clearly and explicitly answers both' because no independent what-statement exists.

Trigger Term Quality

Natural terms a user with this need would actually say are comprehensively present: "Redis", "cache", "cache invalidation strategy", "TTLs", "eviction under memory pressure", "stampede", "stale-while-revalidate", "pub/sub". Not below anchor 4 because no common variation of the request is missing; the terms are the exact vocabulary of this domain, including advanced synonyms.

Distinctiveness Conflict Risk

Clear niche with distinct triggers: every clause is Redis-scoped ("Redis is the app's cache"), separating it from sibling CDN/Varnish cache-testing skills and generic cache-design guidance. Conflict risk is minimal, matching anchor 5 rather than anchor 4, because the distinguishing keyword appears inside the trigger clause itself.