Concepts
Policies, Approvals, And Audit Trails
botYguard policies define what an agent can do before actions reach connected tools. Approvals add human or system review for sensitive operations. Audit trails preserve evidence of requests, decisions, approvals, and outcomes.
Policies
Policies should start narrow. A safe first policy might allow read-only Google Drive access to a specific scope, audit every action, and deny write or delete actions until the workflow is trusted.
Policies can target:
- an agent type, for shared defaults across similar agents
- a specific agent, for narrower control or testing
- an integration, resource, action, and effect
Approvals
Approvals are useful when an action is legitimate but should not run automatically. Examples include write access, deletion, sharing files, sending messages, or using sensitive credentials.
When approval is required, botYguard pauses the operation until the configured review path decides whether the action should continue.
Audit Trails
Audit visibility helps security, platform, and operations teams answer:
- what action the agent requested
- which policy evaluated the request
- whether the action was allowed, denied, audited, redacted, or sent for approval
- who approved sensitive work
- which provider, secret, integration, or tool was used
Review Checklist
- Confirm the policy scope is as narrow as practical.
- Require approval for write, delete, share, or high-impact actions.
- Confirm secrets are passed only through approved execution paths.
- Review audit logs after the first successful workflow.
- Expand access deliberately after the expected behavior is proven.