handoff
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input to generate instructions for subsequent agent sessions, creating a surface for indirect prompt injection.
- Ingestion points: User-provided arguments describing the focus of the next session are ingested via the
argument-hintmechanism. - Boundary markers: Absent. The instructions do not define delimiters or provide specific warnings for the agent to ignore commands potentially embedded within the user input.
- Capability inventory: The skill performs file writing to the user's operating system temporary directory, which is outside the standard workspace.
- Sanitization: The skill relies on a natural language instruction to "Redact any sensitive information" without programmatic validation or escaping of the input data.
- [INDIRECT_PROMPT_INJECTION]: The skill influences the behavior of future agent sessions by generating a "suggested skills" section, establishing a multi-step instruction chain that could be exploited if the handoff document is poisoned.
Audit Metadata