AI agent governance, in plain terms
If you're building or piloting an agent and haven't thought about governance yet, this page is for you.
Governing an agent means deciding, in advance, what it's allowed to do, and building a check that catches it if it goes outside that. Not a policy document. A rule the agent actually runs under.
This matters from the first agent you ship, not once you have many of them. Risk comes from what an agent can do, not from how many you're running.
An agent gets retired the way most projects end: the budget line closes, the service account stays exactly where it was. Nobody revokes what it could still read, write, or trigger.
When an agent's decision gets questioned, there's nothing to point to. Not because the decision was wrong, but because nobody can show what it was based on.
The agent can take a high-impact action end to end, and the first human to see it is whoever deals with the fallout.
None of this requires a large team or a formal compliance function. It requires deciding the rule before the agent needs it, not after something goes wrong.
This library exists to save you from writing those rules from scratch. Each pack is a governance rule for one specific situation, already thought through and tested, built on an open standard rather than one vendor's platform.