desktop-computer-automation

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

• Third-party content exposure detected (high risk: 1.00). The skill requires the agent to take and read screenshots of desktop apps (including browser content) as part of its mandatory screenshot-analyze-act loop (e.g., "search for the weather in Shanghai using the Chrome browser") and also supports ingesting external images via HTTP URLs (example githubassets URL in the tap --locate docs), so it will ingest and act on arbitrary public web/user-provided content visible on-screen or supplied by URL.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill instructs running "npx @midscene/computer@1", which will fetch and execute remote package code from the npm registry (e.g. https://registry.npmjs.org/@midscene/computer) at runtime and is a required dependency for the skill to operate.