gemini-deep-research
Pass
Audited by Gen Agent Trust Hub on Apr 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill facilitates the execution of local TypeScript scripts (e.g.,
scripts/research.ts) using thenpx tsxandbun runrunners. These scripts are used to automate browser interactions and monitor research progress. - [EXTERNAL_DOWNLOADS]: Documentation recommends the installation of the
playwright-corepackage from the standard NPM registry to enable browser automation capabilities. - [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because it ingests and processes large-scale research reports generated by an external AI service (Gemini).
- Ingestion points: Research reports and plan text extracted from
gemini.google.comvia browser automation. - Boundary markers: None identified in the provided documentation to delimit external content from agent instructions.
- Capability inventory: The agent is granted tools for
Bashexecution, file reading (Read), writing (Write), and searching (Glob,Grep). - Sanitization: No sanitization or validation logic is mentioned for the content retrieved from the browser sessions.
Audit Metadata