claude-handoff
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides a specific template for the agent to execute shell commands to manage background processes.
- [INDIRECT_PROMPT_INJECTION]: The skill establishes a mechanism where potentially untrusted data from the current session is passed as instructions to a new agent instance.
- Ingestion points: The handoff summary is constructed from the current conversation history and user-provided arguments specified in SKILL.md.
- Boundary markers: None are specified for the transition between the current session and the new background agent prompt.
- Capability inventory: The skill utilizes the claude CLI to spawn background agents with full access to the working directory.
- Sanitization: While the skill explicitly tells the agent to strip credentials (API keys, passwords), it does not provide instructions for escaping shell metacharacters or preventing instructions in the summary from overriding the new agent's behavior.
Audit Metadata