Governance Guardrail · track

🤖 Governing AI agents

Agents are AI that can act: send email, change records, run code, spend money. They need everything any AI system needs, plus a few things only agents need. Six short lessons, in the order to put them in place. For the technical security of agents, see the CSO Shield.

The agents track is being reviewed — check back soon.

✅ Build an agent approval checklist

Identity, permissions, approval points, logging, kill switch and autonomy stage, in one sign-off.

🗂️ Add your agents to the register

An owner, what each agent can do, and a review date for every one.