rodney
Pass
Audited by Gen Agent Trust Hub on May 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes the
rodneybinary to automate browser tasks. It includes capabilities for evaluating JavaScript in the browser (rodney js), managing browser sessions, and capturing screenshots or PDFs. - [EXTERNAL_DOWNLOADS]: The installation guide suggests downloading the
rodneytool from a public GitHub repository (simonw/rodney) and using package managers likeuvorpipx. These are standard methods for installing developer utilities. - [PROMPT_INJECTION]: This skill has a surface for indirect prompt injection through its web scraping features.
- Ingestion points: Content is retrieved from remote websites using
rodney open,rodney text, androdney html. - Boundary markers: The provided examples do not use specific delimiters or instructions to ignore embedded prompts in the scraped data.
- Capability inventory: The skill can interact with web forms, upload/download files, and run JavaScript, which could be exploited if a malicious page is visited.
- Sanitization: Content from the web is processed without explicit sanitization or filtering before being presented to the agent context.
- [DATA_EXFILTRATION]: The
rodney fileandrodney downloadcommands enable file transfers between the local system and web services. While these are legitimate features, they could be misused to exfiltrate sensitive local files if the agent is directed to do so by untrusted content.
Audit Metadata