game-changing-features
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to 'Research the codebase' to understand the product's current value and identify 10x opportunities. This creates a surface where malicious instructions hidden in the code or documentation of the analyzed project could influence the agent's strategic recommendations.
- Ingestion points: The agent is directed to read and understand the local codebase and existing features in the 'Understand Current Value' step (SKILL.md).
- Boundary markers: The instructions lack explicit delimiters or 'ignore' directives that would help the agent distinguish between code/data to be analyzed and potential instructions embedded within that data.
- Capability inventory: The agent possesses file-read access to the project and file-write access to generate reports in the
.claude/docs/ai/directory. - Sanitization: There are no mechanisms described to sanitize or filter out natural language instructions found in the analyzed files before they are processed by the agent's logic.
Audit Metadata