Lean body that frames only the resolver-tree mechanics specific to this skill rather than generic GraphQL concepts, and is dense with classification tables and executable code. Every section earns its place; not 2 because there is no padded explanation Claude would already know.

Provides fully executable TypeScript for all three fixes (DataLoader, eager projection, selection-set-aware prefetch) plus a concrete jest call-count test and Prisma logging config. Examples are copy-paste ready, matching the anchor.

Steps 1–4 are clearly sequenced with Step 4 an explicit validation checkpoint (call-count assertion) and a DB-layer query-log feedback loop. The classification table functions as a checklist, hitting the anchor for explicit validation steps.

SKILL.md is the overview; deep detail (batch-function contracts, per-request scoping failure, cartesian explosion) lives in references/fixes.md and worked examples in references/examples.md — both verified to exist, one level deep, and clearly signaled with markdown links.

Lists multiple concrete actions — 'Traces a GraphQL resolver tree to locate the N+1 pattern', 'classifies every child field resolver as safe or N+1 risk', and 'applies one of three fixes: per-request DataLoader batching, eager projection... or selection-set-aware prefetch'. This exceeds the anchor for naming the domain and some actions.

Clearly answers both what (trace, classify, fix) and when via an explicit 'Use when...' clause with three concrete trigger conditions. Not 2 because the trigger guidance is explicit rather than implied.

Natural practitioner terms a user would actually say are well covered: 'list-returning resolver', 'connection-pool exhaustion', 'slow-query alert', 'GraphQL traffic', and 'resolver trace shows a child field resolved once per parent row'.

The GraphQL N+1 remediation niche is narrow with distinct triggers unlikely to collide with other skills. Third-person voice ('Traces', 'classifies', 'applies') incurs no voice penalty.