deep-research
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill contains no malicious code, remote execution patterns, or unauthorized access attempts. It serves as an instructional framework for enhancing the agent's internal research processes.
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to systematically fetch and process external web content using the
web_fetchtool. While this increases exposure to potential indirect prompt injection from untrusted websites, it is the primary intended function of the skill. - Ingestion points: External web content fetched via the
web_fetchtool during Phase 2 of the research methodology (SKILL.md). - Boundary markers: No specific delimiters or warnings for the agent to ignore instructions within external content are present.
- Capability inventory: The skill utilizes network-enabled search and fetching tools to retrieve data.
- Sanitization: The skill does not define methods for sanitizing or validating the retrieved information prior to processing.
Audit Metadata