cxas-loss-analysis

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s required runtime workflow fetches and reads CCAI Insights conversation transcripts via fetch_losses.py (calling insights_client.list_conversations(..., view="FULL"), extracting segment text, then SKILL.md Step 3 reads {output_dir}/raw_losses.json and its transcript chunk files for LLM analysis), and those transcripts are outsider-authored (end-customer free text).