handoff
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes current conversation history, which contains untrusted user input, to generate a handoff document for a future agent session, creating a multi-step injection surface. Ingestion points: The conversation context is used as the source for the summary (SKILL.md). Boundary markers: No specific delimiters or warnings are defined for the generated summary content. Capability inventory: The skill involves writing to the file system (temporary directory) and recommending tool calls for the next agent. Sanitization: The skill mandates redaction of credentials and PII but does not include measures to filter malicious natural language instructions.
- [DATA_EXFILTRATION]: The skill instructs the agent to save conversation data to the operating system's temporary directory, which is located outside the managed workspace environment.
Audit Metadata