macos-computer-use

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly instructs the agent to take and "use the window screenshot" to find targets and decide clicks/typing (SKILL.md Fast Workflow step 4) and examples show capturing app windows like Slack (references/examples.md), so the agent will ingest and act on arbitrary third-party/user-generated content displayed in windows.