What it does
A control plane for AI coding agents that makes an agent plan, gather evidence and pass an adversarial review before it gets write access to the workspace. The gates are enforced by the software, not by prompt wording. Each run keeps a sealed verification record of what was checked.
Use cases
- 01Require evidence before an agent edits a repository
- 02Run an adversarial review of a proposed change
- 03Keep a sealed record of what was verified