agent-gate: A deterministic policy gate for AI agent actions

Short summary

agent-gate is a lightweight Python library that inserts a deterministic policy gate between LLM output and real-world actions. It issues single-use capability tokens only after explicit code-level checks pass, treats observed data as non-instructional, backs up files before writes, and maintains a tamper-evident audit log. High-risk actions require explicit human approval by design.

• •Deterministic policy gate sits between LLM decisions and real actions
• •Single-use tokens bound to request hashes prevent replay attacks
• •Sandbox backups, prompt-injection rejection, and tamper-evident audit logging included

Generated with AI, which can make mistakes.