browser-cli
Pass
Audited by Gen Agent Trust Hub on Mar 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes a custom CLI tool,
browser-cli, to execute arbitrary JavaScript code and browser interaction commands within a Firefox environment. - [DATA_EXFILTRATION]: The skill contains functions that write data to the local filesystem. Specifically, the
download()function saves files to the~/Downloads/directory, and theshot()function allows saving screenshots to arbitrary paths provided by the user or agent, such as/tmp/page.png. - [PROMPT_INJECTION]: The skill facilitates an indirect prompt injection surface by retrieving untrusted content from the web and presenting it to the agent.
- Ingestion points: The
read(),snap(), andlogs()functions extract text, HTML structure, and console output from external websites into the agent's context (SKILL.md). - Boundary markers: The skill does not define specific delimiters or warnings to help the agent distinguish between legitimate page content and malicious instructions embedded in web data.
- Capability inventory: The tool possesses capabilities including navigation (
tab), element interaction (click,type), file downloads (download), and filesystem writes (shot) across all scripts (SKILL.md). - Sanitization: There is no evidence of sanitization or filtering of the ingested web content before it is processed by the agent.
Audit Metadata