Conciseness

The body is dominated by useful code, but it re-explains things Claude already knows — the Overview recaps EXPIRE/PEXPIRE/EXPIREAT and the Redis 7 NX/XX/GT/LT flags, and a 'Parsing results' section explains that Redis returns ints, strings, or None and what TTL -1/-2 mean. The redis.io EXPIRE doc URL is also cited three separate times (Overview, 'Basic TTL tests', References). This is 'mostly efficient but includes some unnecessary explanation or could be tightened' rather than the minor-trim level of 4.

Actionability

Most patterns are copy-paste ready against a real fixture — the testcontainers fixture, TTL, conditional-expire, eviction, stampede, and pubsub tests are complete and executable, with a runnable pytest command and CI service config. It misses 5 because two key examples depend on undefined application-side helpers: 'test_write_invalidates_cache(r, db)' calls an undefined 'update_user_via_app' and 'test_tenant_namespaced_keys(r, cache)' uses an undefined 'cache' object, so those blocks cannot run as written.

Workflow Clarity

The skill is a pattern catalog with a loose install → fixture → write tests → run → CI progression, but it never states that sequence, and there are no validation checkpoints between steps. Sequencing is also muddled: the 'Pub-sub invalidation across nodes' test pattern is nested under '## Running' after the pytest instructions, separating it from the other test patterns in '## Authoring'. This is 'steps/patterns listed but validation gaps; checkpoints missing or implicit'.

Progressive Disclosure

SKILL.md links to only 2 of the 6 bundle files (stampede.md and cache-coherence-patterns.md); the other four (stale-while-revalidate.md, stampede-mitigations.md, rfc-9111-http-caching-directives.md, cross-tier-coherence-and-test-surface.md) are reachable only through a second hop via cache-coherence-patterns.md, which labels itself a 'Deep reference' — i.e., references are two levels deep and not directly signaled from SKILL.md. The stale-while-revalidate file is especially orphaned given the description promises that pattern. This fits 'references present but not clearly signaled / could be better organized' rather than the minor-gap level of 4.

Specificity

The description enumerates several concrete capability 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' — which matches the 'several specific actions; minor gaps' anchor. It falls short of 5 because the actual action (writing tests for these behaviors) is never stated as an explicit verb; the pattern list implies it rather than names it.

Completeness

The 'when' is explicit and strong ('Use when Redis is the app's cache and its behavior needs tests, or when choosing a cache invalidation strategy'), and the enumerated pattern list conveys the 'what', but the what is never expressed as a concrete action (e.g., 'Writes tests for...'). This matches anchor 4 ('both present; one could be more explicit') rather than anchor 5, which requires an explicitly stated capability paired with the trigger clause.

Trigger Term Quality

Covers the natural vocabulary a user would invoke when needing this skill — 'Redis', 'cache', 'cache invalidation', 'TTLs', 'eviction', 'stampede', 'stale-while-revalidate', 'pub/sub', 'cache-aside' — including both the informal terms (Redis, cache) and the technical synonyms (TTL, eviction, SWR). This is comprehensive trigger coverage for the niche; nothing common is missing.

Distinctiveness Conflict Risk

The Redis-as-app-cache testing niche is clear and distinct from generic cache or database skills, but the clause 'or when choosing a cache invalidation strategy' deliberately broadens the skill beyond testing into strategy selection, creating minor overlap risk with sibling caching skills (e.g., the cdn-cache-purge-tests skill it itself references). This is 'mostly distinct; minor overlap risk' rather than a fully clean 5.