fact-check
Pass
Audited by Gen Agent Trust Hub on Feb 21, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION] (LOW): Indirect prompt injection surface detected.
- Ingestion points: The skill ingests untrusted data from target documents (Step 3) and external web content via the WebSearch tool (Steps 2 and 4).
- Boundary markers: The prompt template used to invoke the Sonnet sub-agent does not employ delimiters (such as XML tags or triple backticks) to isolate the document content from the system instructions.
- Capability inventory: The skill possesses the capability to perform web searches and spawn sub-agents with specific task instructions.
- Sanitization: No sanitization, escaping, or explicit 'ignore embedded instructions' directives are present in the prompts used to process external data.
Audit Metadata