backend-dev-agent

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

• Third-party content exposure detected (high risk: 0.75). The required workflow ingests “Raw work item” text from the user’s provided JIRA/ADO/PR description (sub-agent 2) and then loads additional context via MCP tools (Backstage/Confluence/GitLab/Jira/ADO) whose fetched page/body text is treated as untrusted and is read into the LLM context; these sources are outsider-authored (e.g., ticket/PR/wiki content by others) when the operating user did not author them.