claude-handoff
Pass
Audited by Gen Agent Trust Hub on Jul 9, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The agent is instructed to launch background processes using the
claudecommand-line tool. - Evidence: Template string
claude --bg --name "<descriptive name>" "<handoff summary>"inSKILL.md. - [DATA_EXFILTRATION]: The instructions suggest referencing external resources via URL in the handoff documentation.
- Evidence: "Reference them by path or URL instead" in
SKILL.md. - [INDIRECT_PROMPT_INJECTION]: The skill acts as a bridge for conversation state between agents, creating an attack surface for instructions embedded in history to influence subsequent sessions.
- Ingestion points: The entire conversation history and user-provided arguments are used to generate the summary.
- Boundary markers: No specific delimiters are required for the handoff summary content.
- Capability inventory: Shell command execution via the
claude --bgcommand. - Sanitization: The skill provides specific instructions to redact sensitive credentials and PII.
Audit Metadata