handoff
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests the current conversation history, which may contain untrusted data, to generate a handoff summary. This creates a surface where embedded instructions in the history could influence the output.
- Ingestion points: The agent is instructed to summarize the 'current conversation' in
SKILL.md. - Boundary markers: There are no explicit delimiters or markers defined for the generated document to isolate summarized content from the original context.
- Capability inventory: The skill performs a file-write operation to the user's OS temporary directory to store the handoff document.
- Sanitization: The instructions include a specific requirement to redact sensitive information such as API keys, passwords, and PII, which acts as a manual sanitization step for data exfiltration risks.
Audit Metadata