skill-iter-tune

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The orchestrator routinely sends full skill files and produced artifacts to external LLM endpoints (ccw cli → Claude/Gemini) and runs a general-purpose Agent that can automatically modify local skill files, which together create clear high-risk patterns for data exfiltration and for introducing backdoors or malicious changes (supply‑chain abuse), plus the potential to leak credentials if any secrets are present in the project files.