llm-safe-haven

A published npm tool: one command to fasten the security bolts on AI coding agents, aimed at solo developers who run agents with broad permissions and no security team behind them.

• One command instead of a hardening checklist nobody finishes.
• Ships agent hooks and manifests -- the guardrails live where the agent runs, not in a wiki.

Where to look, in the repo:

1. README.md -- what it hardens and the one-liner install.
2. hooks/ -- the agent-side guardrails.
3. SECURITY.md -- the threat model it works against.