Pillar 04 · Accountability
C3 Agent Charter
Write down what the agent may decide before it decides anything.
Form Document
What it's for
An agent that sends, changes or approves things without a person checking each action carries more risk than one that drafts for review. The charter is the business agreement for one agent: what it is for, what it may do on its own, what it must hand to a person, what it must never do, who checks its work and how to stop it. The technical controls that enforce it live in the agent's permission record (A4).
Every agent rated Medium or High in the use-case register (S1) needs a charter signed before go-live. For Low-tier agents it is good practice.
What you get
- Decision boundaries What it may do on its own, what it must hand to a person and what it must never do.
- Human oversight Who checks its work, what they check and what they may do about it.
- Stopping it Who can stop it, how and when, and the last stop test.
- Change and sign-off Re-approval when its model, sources or permissions change.
How it connects
One charter per S1 agent; its technical limits live in A4's permission record, its model is checked in S4, and its owners and reviewer come from C1 and C2.
A flavour of the worked example
| What it is for | Draft the quarterly commentary for each fund from certified portfolio metrics and the reporting policies |
|---|---|
| May do on its own | Read certified metrics (S-004) and reporting policies (S-003) |
| Must hand to a person | Any figure that isn't a certified term (M1) |
| Must never do | Send anything outside the firm |
| Who can stop it | Agent owner, use-case owner, governance office or security, without asking anyone |
A few fields from one record of the toolkit's worked example, with long text shortened.
Next step
Guided build coming next.
Request access
Ask for access to the guided builds, including this kit's when it is ready.