LLM behavioral enforcement for agent systems: rules preventing context loss, file damage, drift.
Verification confirms publisher identity (repo ownership), not code safety. The security scan covers known CVEs and suspicious install scripts — it cannot prove the absence of malicious code.
LLM behavioral enforcement for agent systems: rules preventing context loss, file damage, drift.