agent-browser

Pass

Audited by Gen Agent Trust Hub on Mar 18, 2026

Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the infsh (inference.sh) CLI to manage browser sessions and perform automation tasks.
  • [REMOTE_CODE_EXECUTION]: The execute function allows the agent to run arbitrary JavaScript code within the browser context, which is a powerful capability that could be misused if the agent is influenced by malicious input.
  • [DATA_EXFILTRATION]: The skill provides tools for extracting browser cookies and uploading local files to remote websites, creating a pathway for potential exfiltration of sensitive information.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted content from the web.
  • Ingestion points: Untrusted data from web pages is brought into the agent's context through the open, snapshot, and execute functions (documented in SKILL.md and references/commands.md).
  • Boundary markers: The skill does not implement boundary markers or instructions for the agent to ignore embedded commands within the scraped web content.
  • Capability inventory: The skill possesses capabilities for arbitrary JavaScript execution, file uploads, and media capture (screenshots/video), which could be leveraged by an indirect injection attack.
  • Sanitization: There is no evidence of sanitization or filtering of the HTML or text content retrieved from websites before it is provided to the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 18, 2026, 01:25 AM