canvas-lms

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

• Third-party content exposure detected (high risk: 0.75). The required runtime workflow uses the Canvas MCP server tools to call Canvas APIs (e.g., to read discussion topics, conversations, pages, files, submissions, and reports), and those API responses can include outsider-authored free text (students/other instructors’ messages, discussion posts, wiki/page HTML, submission comments) that the agent will ingest into LLM context.