simplify
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill identifies and processes recently modified code blocks, creating a surface for instructions embedded within that code to influence the agent's behavior.\n
- Ingestion points: Recently modified code sections in the current session (SKILL.md).\n
- Boundary markers: Absent. The instructions do not define clear delimiters or specific commands to ignore potential instructions within the analyzed code.\n
- Capability inventory: The agent performs file modification and code generation based on the analysis (SKILL.md).\n
- Sanitization: Absent. There is no evidence of logic to filter or escape instructions embedded in the input code.
Audit Metadata