agent-reflex
rutvikchandla3/agent-reflex
A ~400ms System-One safety reflex for AI coding agents: auto-approves routine tool calls, escalates risky ones, blocks exfiltration & prompt injection. Powered by TypeSafe's Jev.
TypeScript
- Stars
- 0
- Forks
- 0
Why it is listed
Coding-agent permission layer uses Jev risk and context signals with deterministic allow/ask/deny policy; unreachable Jev defers to the agent’s normal permission behavior.