handoff
Pass
Audited by Gen Agent Trust Hub on Jun 16, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted conversation data from the current session to generate summaries.
- Ingestion points: Summarizes the "current conversation" (SKILL.md).
- Boundary markers: The skill does not define boundary markers or delimiters to isolate the summarized content from instructions that might be embedded within the conversation history.
- Capability inventory: The skill utilizes the agent's capability to write files to the operating system's temporary directory.
- Sanitization: The skill explicitly includes a directive to redact sensitive information (API keys, passwords, PII) before finalizing the document.
- [COMMAND_EXECUTION]: The instructions require the agent to write a document to the system's temporary directory, which involves file system operations outside of the immediate project workspace.
Audit Metadata