Loading…
Pattern library
The same risky actions, in every company.
A pattern is a shape of risky execution that recurs regardless of industry — an action that spends money nobody approved, a deploy that assumes a review it never had, an agent that reaches data it should not keep. 15 patterns, each with the signals that decide it and the 53 executable policies that implement them.
Protecting money
Payments, payouts, refunds, spend and treasury movement.
Prevent AI overspending
Cap what an agent can spend before the spend happens, not in next month's bill.
6 executable policies
Treasury payment release
Stop the payment that clears sanctions on paper but not in fact.
7 executable policies
Customer refund
Let support move fast without letting policy drift.
5 executable policies
Cloud spend increase
Commit infrastructure budget deliberately, not by autoscale.
4 executable policies
Payroll and HR process release
Check the run before it commits, not in the reconciliation.
3 executable policies
Protecting code
Deploys, migrations, branches and release gates.
Protecting infrastructure
Production data, cloud resources, secrets and network egress.
Control MCP egress
Decide what leaves before it leaves, at the tool-call boundary.
8 executable policies
Prevent destructive production changes
Make deleting production a decision, not a command.
3 executable policies
Third-party data sharing
Decide what reaches an external system before the request is made.
6 executable policies
Protecting ai agents
Tool calls, retrieval, memory and autonomous action.
Protecting approvals
Delegated authority, segregation of duties and escalation paths.
Protecting commerce
Checkout, pricing, discounts, inventory and fulfilment.
Protecting identity
Access grants, privilege, offboarding and who may act for whom.
Browse by concept instead
Each pattern belongs to one or more concepts. If you arrived with a term rather than a workflow, start there.