qa-automation

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

• Third-party content exposure detected (high risk: 0.78). Notion/Plane inputs are treated as active sources: when Notion MCP/Plane MCP is used, the skill reads outsider-authored page/database/comment text (e.g., Notion test case fields or Plane work item descriptions/comments) and incorporates that readable prose into the agent’s LLM context during “Resolve Notion and Plane source refs” and “Resolve Notion and Plane write policies before any write.”

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly treats Notion and Plane URLs (e.g., https://notion.so/workspace/Auth-Test-Cases-db123 and https://app.plane.so/acme/projects/dki/issues/DKI-179) as runtime inputs to be read and used to drive automation mapping, status updates and sync actions, so fetched external content can directly influence agent instructions and behavior.