cmux

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt includes commands that read and return sensitive values (e.g., eval "localStorage.getItem('token')", get value, and an example filling a plaintext password "secret123"), which would require the agent to capture and could output secrets verbatim.

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

• Third-party content exposure detected (high risk: 1.00). The SKILL.md "Browser Surfaces" section shows cmux can open arbitrary URLs (cmux browser open), read page content (get text/html, snapshot, get attr), execute/eval JavaScript and inject scripts/styles (eval, addscript, addstyle), and then make navigation/click decisions based on that content, meaning it ingests untrusted public web content that can materially influence actions.