agent-browser

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS] (SAFE): The skill installs the agent-browser package via NPM and downloads the Chromium browser. Since the developer (Vercel) is a trusted organization, these downloads are considered safe per the [TRUST-SCOPE-RULE].
  • [COMMAND_EXECUTION] (SAFE): The skill executes the agent-browser CLI via Bash to perform web interactions. All commands are standard for the tool's purpose, with no evidence of malicious command concatenation or shell injection patterns.
  • [INDIRECT_PROMPT_INJECTION] (LOW): The skill possesses a surface for indirect prompt injection as it ingests and processes untrusted web content.
  • Ingestion points: Web content is ingested through agent-browser snapshot and agent-browser get text commands in SKILL.md.
  • Boundary markers: Absent. There are no explicit instructions or delimiters provided to help the agent distinguish between tool output and malicious instructions embedded in web pages.
  • Capability inventory: The skill can perform file-system writes (screenshot, pdf), web interactions (click, fill, select), and navigation (open).
  • Sanitization: No sanitization or content filtering is implemented for the data retrieved from external URLs.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:30 PM