browser-automation
Pass
Audited by Gen Agent Trust Hub on Apr 3, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches and executes the
@midscene/webpackage from the npm registry using thenpxcommand to perform browser automation tasks. - [COMMAND_EXECUTION]: Utilizes shell commands to run the automation framework and performs connectivity prechecks using
curlagainst local browser debugging ports (127.0.0.1:9222 and 127.0.0.1:3766). - [PROMPT_INJECTION]: The skill has a surface area for indirect prompt injection as it processes and reacts to content from external websites.
- Ingestion points: External web content and screenshots are ingested via the
connect --urlcommand (File: SKILL.md). - Boundary markers: Absent; the multimodal model interprets the entire visual state of the browser without specific markers to distinguish page content from instructions.
- Capability inventory: Full browser interaction including clicking, typing, and navigation via the
actandtaptools (File: SKILL.md). - Sanitization: None; the skill relies on the underlying model's ability to distinguish between page elements and instructions.
Audit Metadata