claude-handoff

Pass

Audited by Gen Agent Trust Hub on Oct 7, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill uses summarized conversation history to seed the prompt of a new background agent session.\n
  • Ingestion points: The current conversation history is used as the source for the handoff summary.\n
  • Boundary markers: There are no explicit markers or delimiters defined for the summary text within the command line arguments.\n
  • Capability inventory: The background agent launched by the CLI typically possesses a wide range of tool-use capabilities.\n
  • Sanitization: The instructions include a specific directive to redact sensitive information such as API keys and PII, providing a level of manual sanitization.\n- [COMMAND_EXECUTION]: The skill instructs the agent to construct and execute a shell command (claude --bg --name \"<name>\" \"<summary>\"). This involves dynamic argument construction based on conversation content, which requires the agent to handle shell escaping correctly.\n- [DATA_EXFILTRATION]: The process involves transferring conversation context, which may contain sensitive information, to a new agent session prompt. The skill includes instructions to redact credentials and PII to mitigate data exposure during this transfer.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 7, 2026, 11:14 AM