skill-forge

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The code intentionally collects user chat transcripts as example messages and includes them in PR bodies that are pushed/created to a remote GitHub repository, which is an explicit data-exfiltration capability (uploads local transcript content to an external repo).

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

• Third-party content exposure detected (medium risk: 0.30). In audit mode, scripts/mine_chats.py ingests outsider-authored free text from user-submitted Claude Code chat transcripts in ~/.claude/projects/**/*.jsonl (extract_user_text → topic/painpoint extraction → reports/latest.{md,json}).