atypica-user-interview

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

• Third-party content exposure detected (high risk: 0.80). The skill fetches and ingests public content from atypica.ai (e.g., atypica_universal_search_personas / atypica_universal_get_persona which returns the persona "prompt" and atypica_universal_get_report which returns HTML shareable reports), and those persona prompts/public reports are read and used by the agent to drive interviews and decisions, enabling untrusted user-generated content to influence agent behavior.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly calls the MCP API at https://atypica.ai/mcp/universal (via callTool examples and scripts/mcp-call.sh) at runtime to fetch persona data (e.g., atypica_universal_get_persona returning persona.prompt) which directly injects prompt/instruction content that controls agent behavior, and the skill depends on that endpoint to operate.