claude-handoff
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the
claudecommand-line interface to launch background processes. This is the primary function of the skill and uses standard CLI arguments to pass descriptive names and summaries. - [INDIRECT_PROMPT_INJECTION]: The skill generates a summary based on conversation history and user arguments, which is then used as the initial prompt for a new agent session. This creates a surface where external data could potentially influence the behavior of the next agent.
- Ingestion points: Current conversation history and user-supplied arguments are used to generate the summary (SKILL.md).
- Boundary markers: The summary is passed to the CLI as a quoted string, but the skill does not specify instruction-level delimiters within the summary itself.
- Capability inventory: Executes shell commands via the
claudeCLI (SKILL.md). - Sanitization: The skill includes a specific instruction to redact sensitive information such as API keys, passwords, and PII before the summary is generated.
Audit Metadata