canvas-exam-schedule-compiler

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

• Third-party content exposure detected (high risk: 0.75). The skill ingests Canvas course announcement message fields (outsider-authored course content by instructors/other users) via canvas-canvas_list_announcements, which are free-text and can contain indirect prompt-injection payloads into the agent’s LLM context.