deep-research-pro
Warn
Audited by Gen Agent Trust Hub on Apr 16, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The README.md file instructs users to clone a repository from an unverified GitHub account (https://github.com/parags/deep-research-pro.git) and execute a script (scripts/research) that automatically downloads and installs dependencies from the internet using the uv package manager.
- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as described in SKILL.md. It fetches full-page content from arbitrary URLs via web_fetch and synthesizes this data without using boundary markers or specific instructions to ignore embedded malicious content.
- Ingestion points: Content retrieved via web_fetch in Step 4 of the workflow.
- Boundary markers: None identified in the workflow instructions.
- Capability inventory: File system writes to the ~/research/ directory and the ability to spawn subagents using sessions_spawn.
- Sanitization: No sanitization or validation of the fetched web content is performed before processing.
- [COMMAND_EXECUTION]: The documentation encourages the execution of a local script (scripts/research) that performs complex tasks including environment management and multi-query execution, which represents a risk if the externally sourced script contains malicious logic.
Audit Metadata