scenario-refiner

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

• Third-party content exposure detected (high risk: 0.85). Runtime path: the workflow fetches outsider-authored conversation text by calling ttai:get_sessions_batch and then downloading each transcript_url contents (public/third-party session transcripts) and quoting them into the agent’s LLM context for diagnosis.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill uses the external MCP API at https://api.toughtongueai.com/api/public/mcp at runtime to fetch scenario fields (ai_instructions, strategy, tools_config and transcript contents) which are injected into the compiled system prompt, so remote content directly controls the agent's prompts.