tailrocks-graphql-best-practices
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process and evolve untrusted repository content, including SDL snapshots, source code, and tool outputs. This creates a surface for indirect prompt injection.
- Ingestion points: Identified in
references/runtime-trust.mdas repository files, reports, fixtures, scripts, registry content, and web content. - Boundary markers: The skill includes a dedicated
runtime-trust.mdreference that explicitly instructs the agent to treat all repository data as untrusted and prevents it from altering scope or authority. - Capability inventory: The skill involves writing and modifying source code and schemas, and executing development tools like
cargoandbunvia subprocess calls. - Sanitization: The skill enforces contract integrity through mandatory schema snapshotting and breaking-change diffs using
graphql-inspectoras detailed inreferences/contract-gates.md. - [COMMAND_EXECUTION]: The instructions involve executing standard development and build tools, including
cargofor Rust compilation/schema generation andbunfor TypeScript codegen and contract gating. These commands are standard for the intended use case of evolving a GraphQL service and are scoped to the local project environment.
Audit Metadata