visual-explainer

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

• Third-party content exposure detected (medium risk: 0.30). The visual-explainer workflow reads and ingests user/outsider-provided free-text indirectly via the orchestrator passing untrusted SOURCE/brief fields into sub-agents (e.g., ve-diagram-builder input contract) to generate MDX/HTML fragments.

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

• Potentially malicious external URL detected (high risk: 0.90). SKILL.md explicitly instructs at runtime to clone and install https://github.com/theclaymethod/artifacture (clone then npm install and run npm scripts), which fetches remote code that will be executed and is a required runtime dependency.