handoff
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted conversation data to create a summary that guides future agent behavior, potentially carrying over malicious instructions.\n
- Ingestion points: The current conversation history (SKILL.md).\n
- Boundary markers: The instructions lack delimiters to isolate conversation data from the summary generation logic.\n
- Capability inventory: The skill specifies writing files to the system temporary directory outside the project workspace.\n
- Sanitization: The prompt explicitly directs the agent to redact credentials and PII, which serves as a security control against data exposure.
Audit Metadata