accessibility-skill

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

• Third-party content exposure detected (high risk: 0.75). The required workflow’s only runtime text input into the agent LLM context is through the Accessibility MCP server tool’s getAccessibilityReport/build+scan flow, which takes a public URL and returns a scan report (i.e., outsider-authored/free-text content from arbitrary external web pages) into the agent/tool output that the LLM consumes.