agent-oversight The Permission Layer Is 98% of Agent Engineering Only 1–2% of agent code is actual AI logic. The other 98% — permission systems, hook composition, sandboxing, context management, subagent delegation — is what determines whether your agent is safe to run in production.
agent-oversight Why Claude Code PreToolUse Hooks Can Still Be Bypassed Your Claude Code hooks can block `cat .env` and still leak your secrets. Here's exactly why — and the four-layer stack that actually bounds blast radius.