Evaluate proposed agent actions against explicit scope, identity, permissions, risk, budget, and approval requirements.
JAHIDS Governor Policy
Use when
An agent proposes a tool call, data access, deployment, publication, or other consequential action.
Procedure
- Identify the requesting principal, tenant, workspace, mission, and agent identity.
- Resolve policy from trusted server-side context; never trust tenant IDs supplied only in headers or tool text.
- Check action, target, environment, data scope, and permission.
- Assess risk, budget, reversibility, and blast radius.
- Require approval for actions outside the granted autonomy level.
- Bind approval to the exact payload hash, target, environment, scope, approver, and expiry.
- Return ALLOW, DENY, or REQUIRE_APPROVAL with a reason and evidence reference.
Output
A structured policy decision and audit evidence.
Authority boundary
A skill cannot grant itself permissions, bypass the Governor, or convert a proposal into authorization.
Related skills
- Downloads
- 61All time
- Security audit
- Pass
- Last updated
- 23h ago
- Current version
- v0.1.0
- License
- MIT-0
