Add compact prompt-defense baselines to active ECC prompt surfaces and copied CLAUDE examples. AgentShield prompt-defense findings are now zero; local tests passed 2366/2366.
- Expand permissions.allow (git, node, npm, npx, ls, cat, Read, Edit, Write, Glob, Grep) to reduce approval prompts - Set CLAUDE_ECC_MODE=standard in env for consistent hook profile - Gate stop-hook-git-check.sh behind ECC_GIT_CHECK=1 to prevent blocking mid-session - Add project-level Node.js rules for ECC repo (stack, file conventions, hook dev standards) - observe.sh confirmed already async — no change needed https://claude.ai/code/session_014SqRA9y6HwVVPGeSCJfwVF