camoufox

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill's examples show form-fill actions that place plaintext credentials into the action JSON (e.g., "value": "secret"/"alice"), which instructs the agent to embed user passwords verbatim in generated requests/actions.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The skill exposes an MCP browser-browsing workflow (e.g., browse, browse_find, browse_links, and session snapshot/screenshot tools) that ingests free text from an outsider-controlled webpage the caller supplies via a url at runtime, so the text is read before any attacker-specific item selection can be enforced by the agent.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill includes runtime npx commands that fetch and execute remote npm packages (e.g., "npx -y camoufox-mcp-server@latest" and "npx -y camoufox-js@0.12.0 fetch"), which are required to run the Camoufox server and therefore execute remote code at runtime.