# agentlock

Canonical URL: https://doramagic.ai/en/projects/agentlock/

Source repository: https://github.com/webpro255/agentlock

## What it is

An adversarially benchmarked reference implementation for pre-action agent authorization. Framework-agnostic tool permissions, identity verification, scoped access control, and audit logging for any AI agent.

## Capability boundary

skill, recipe, host_instruction, eval, preflight

## First safe verification

Verify the smallest path in an isolated environment and keep a rollback path.

## Main risk

May increase setup, validation, or first-run risk for the user.

## Evidence base

https://github.com/webpro255/agentlock, https://github.com/webpro255/agentlock#readme, Human Manual, Pitfall Log