truth-file
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process potentially untrusted external data, such as customer research and marketing analytics, which represents a surface for indirect prompt injection.
- Ingestion points: Reads existing project files including research, analytics, and user-provided context (SKILL.md).
- Boundary markers: The skill specifically instructs the agent to preserve customer wording with its origin and use boundaries, and defines evidence states (verified, user-provided, etc.) in references/context-workflow.md.
- Capability inventory: Performs local file read operations on project documents and write operations to the .agents/ directory.
- Sanitization: Relies on structured labeling of information states to distinguish between verified data and raw user inputs.
Audit Metadata