handoff
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructions direct the agent to 'Read any existing HANDOFF.md file in the current directory and follow its instructions to continue the work.' This creates a vulnerability where an attacker could place malicious instructions in a HANDOFF.md file that the agent would then execute.
- Ingestion points: The agent reads content from HANDOFF.md.
- Boundary markers: Absent; there are no instructions to treat the file content as untrusted.
- Capability inventory: The skill manages files, giving the agent read/write access to the workspace.
- Sanitization: No sanitization of the HANDOFF.md content is performed before the agent follows its instructions.
Audit Metadata