handoff
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-generated conversation history, which is a potential surface for indirect prompt injection (Ingestion point: SKILL.md). The instructions mitigate this by explicitly requiring the agent to redact sensitive information like API keys, passwords, and PII before generating the handoff document (Sanitization: Present). Additionally, the skill configuration disables automatic model invocation, ensuring that state transitions occur only with direct user interaction.
- [COMMAND_EXECUTION]: The skill involves writing a summary document to the user's operating system temporary directory (Capability inventory: File-write). This is a functional requirement for its task and does not involve unauthorized or suspicious file system access.
- [SAFE]: The skill is documented as a port from a public repository on GitHub (github.com/mattpocock/skills) for attribution purposes. The analysis confirms the package is self-contained and does not introduce external dependencies, remote script downloads, or hidden network operations.
Audit Metadata