chrome-debug

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION] (LOW): This skill exposes a surface for indirect prompt injection by allowing the agent to load and process data from arbitrary external URLs. 1. Ingestion points: Untrusted data enters the context via mcp__chrome-devtools__navigate_page and mcp__chrome-devtools__take_snapshot. 2. Boundary markers: None; the skill instructions do not define delimiters or warnings for the agent to ignore embedded instructions in web content. 3. Capability inventory: The agent can execute shell commands (npm run chrome), write files (take_screenshot), and interact with page elements (click, fill). 4. Sanitization: None; no validation or filtering of browser-retrieved data is specified.
  • [COMMAND_EXECUTION] (SAFE): The use of npm run chrome to initialize the browser environment is a standard local operation and does not process external or untrusted strings.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:42 PM