map
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill scans a wide variety of project files to synthesize a domain model and architecture diagram, creating a surface for potential injection from the codebase content.
- Ingestion points: Manifest files (package.json, Cargo.toml, go.mod, pyproject.toml, requirements.txt, mix.exs, Gemfile, composer.json), database schemas (Prisma, SQL migrations, TypeORM/Sequelize, GraphQL), and application route files as specified in the exploration instructions.
- Boundary markers: The instructions define output markers (map:start/end) for the CLAUDE.md file but do not provide delimiters or instructions for the agent to ignore prompt-like content found within the ingested source files.
- Capability inventory: The skill utilizes codebase exploration via a subagent, reads multiple file types across the project, and writes results to the local filesystem (CLAUDE.md).
- Sanitization: No sanitization or verification steps are outlined for the data gathered from the codebase files before it is processed by the agent.
Audit Metadata