plexy-recipe
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes project source code and user-provided snippets to generate persistent documentation (recipes) that the agent uses for future implementation tasks. This creates a surface where malicious instructions inside the source files could potentially influence subsequent agent behavior.
- Ingestion points: Reads provided snippets, project code, and existing recipes in .agents/recipes/.
- Boundary markers: No explicit delimiters are defined to separate analyzed code from instructions, though the skill encourages structural separation in markdown.
- Capability inventory: Reads and writes local Markdown files within the .agents/recipes/ directory.
- Sanitization: The instructions explicitly direct the agent to "Omit irrelevant details and secrets" and "Do not endorse unsafe or accidental behavior as a recommendation."
Audit Metadata