Conciseness

Largely lean, domain-specific guidance (IS/IS-NOT scope, vulnerable vs. secure examples) that earns its tokens, but the 'what it is NOT' list and the Phase-1 gates are restated inside the Phase 2 subagent prompt and again in Important Reminders.

Actionability

Provides copy-paste-ready subagent prompt templates with exact output markdown schemas, a concrete batching procedure (8 sites -> 3 batches), and executable vulnerable/secure code examples for Node.js and Python.

Workflow Clarity

Three phases are explicitly sequenced with Gate 1 and Gate 2 validation checkpoints, parallel batch launch, and intermediate-file cleanup; the batch operation has real validation, so the destructive/batch cap does not apply.

Progressive Disclosure

Well-sectioned but a ~350-line monolithic SKILL.md with no bundle files; the large Phase 1/Phase 2 prompt templates and the vulnerable/secure examples library are inlined where one-level-deep reference files would aid navigation.

Specificity

Lists multiple concrete actions across three named phases (recon, batched verify, merge) with sub-tasks like 'find unsafe operation document assembly sites' and 'trace user input to those sites in parallel subagents, up to 3 candidate sites each'.

Completeness

Explicitly answers both what it does (three-phase GraphQL injection detection with prerequisites and outputs) and when to use it via a clear 'Use when asked to find...' clause with concrete triggers.

Trigger Term Quality

Provides natural trigger phrasings ('GraphQL injection', 'unsafe GraphQL document construction', 'operation string injection bugs') but lacks synonyms or file-extension variants a user might also say.

Distinctiveness Conflict Risk

Occupies a clear niche (operation-document GraphQL injection) and explicitly disambiguates from resolver SQL/NoSQL injection covered by other SAST skills, minimizing wrong-skill triggering.