The body is mostly dense, purposeful reference material (tables, code, numbered flows), but the opening doc-quote definition of what a persisted query is and framing prose like 'Two motivations, often conflated' and 'Best of both: dev gets the iteration speed...' could be tightened. Not score 3 because some explanatory framing explains concepts Claude already knows; not score 1 because the bulk is lean and reference-style rather than padded.

Provides fully executable examples: ApolloServer persistedQueries config, the createPersistedQueryLink TypeScript snippet, and `npx graphql-codegen` commands, plus concrete per-mode request flows. Not score 2 because the core guidance is copy-paste ready rather than pseudocode.

Each mode has a clearly numbered request flow (e.g. Mode 1 steps 1-5, Mode 2 steps 1-5), and the 'Testable behaviours' table plus the 'Anti-patterns' table provide explicit verification and error-recovery checkpoints ('These tests prove the chosen mode is actually in effect'). Not score 2 because validation/verification is explicit, not merely implied.

Well-organized single-file structure with clearly labeled sections (Overview, When to use, the three modes, Implementation patterns, Per-framework support, Testable behaviours, Anti-patterns, Limitations, References) and clearly signaled one-level external links; no nested/deep references. Not score 2 because navigation is clean and content is appropriately sectioned for a self-contained reference with no bundle files.

Lists multiple concrete capabilities in one sentence: 'the Apollo APQ SHA-256 hash protocol, the PersistedQueryNotFoundError retry flow, the `extensions.persistedQuery` payload, GET-vs-POST and CDN-cache implications, and the three operation modes'. Matches the score-3 anchor of listing multiple specific concrete actions; not score 2 because the action set is comprehensive rather than partial.

Answers both what (a pure-reference catalog of specific strategies) and when via the explicit trigger 'Use to design or audit a server's persisted-query request layer'. Not score 2 because the 'when' is explicit, not merely implied.

Uses natural in-domain terms a user would actually say when needing this skill — 'GraphQL Persisted Query', 'Apollo APQ', 'allowlist', 'CDN-cache', 'auto-register'. Not score 2 because common variations are well covered rather than missing.

Clear niche (GraphQL persisted-query strategy) plus active disambiguation: 'This strategy reference emits no tests; to author the runtime tests use apollo-server-tests, graphql-yoga-tests, or mercurius-tests'. Unlikely to trigger for the wrong skill; not score 2 because the boundary is explicit.