deep-research-pro
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to search for and read external web content in full, which creates a surface for indirect prompt injection where malicious instructions embedded in web pages could influence the agent's summary or subsequent actions.
- Ingestion points: Workflow Step 4 in SKILL.md instructs the agent to 'Deep-read the strongest sources' and 'Open and read the most relevant pages in full.'
- Boundary markers: Absent. The instructions do not provide specific delimiters or 'ignore embedded instructions' warnings to isolate external data from the system prompt.
- Capability inventory: The agent uses built-in search and browser tools to fetch external data and is instructed to write reports to the workspace (SKILL.md Step 6).
- Sanitization: Absent. There is no requirement or instruction to filter, validate, or sanitize the content fetched from the web before synthesis.
Audit Metadata