research
Pass
Audited by Gen Agent Trust Hub on Jun 30, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection via the web content it retrieves.
- Ingestion points: The skill utilizes a
WebFetchtool within Phase 2 (Parallel Research Dispatch) to read primary sources, official documents, and articles directly from external URLs into the agent's context. - Boundary markers: The sub-agent prompt templates do not include explicit structural delimiters (e.g., XML tags or clear 'BEGIN/END DATA' blocks) or specific instructions for the agent to ignore potential commands embedded in the fetched text.
- Capability inventory: The skill possesses the capability to write synthesized findings to the local filesystem in the
docs/research/directory, which could be exploited by a malicious website providing deceptive instructions. - Sanitization: The instructions do not describe any sanitization, filtering, or validation process for the content retrieved via web tools before it is processed by the model for analysis.
Audit Metadata