iblai-vibe-agent-chat

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly tells the agent to ask the user for their agent/mentor UUID and write it directly into .env.local using the Edit tool (i.e. embed the user-supplied value in generated file output), which requires the LLM to output a secret/identifier verbatim.

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

• Third-party content exposure detected (medium risk: 0.30). SKILL.md wires an in-process Chat React component from @iblai/iblai-js/web-containers/next that is mounted on app/agents/[mentorId]/chat-new/page.tsx and therefore consumes outsider-authored message input from the host UI at runtime, before any explicit content selection is mentioned.

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

• Potentially malicious external URL detected (high risk: 0.90). The ChatWidget passes a runtime mentor URL to the embedded web component which loads the hosted MentorAI platform (e.g. https://mentorai.iblai.app / https://mentorai.${domain}) into an iframe, so remote content from that URL is fetched at runtime and controls the chat/agent UI and behavior.