Patterns

Patterns are reusable verification structures for agentic systems. Each card names the failure it controls, the mechanism it uses, the observable signal it produces, and the evidence behind it. Two ways in: start from the failure you are seeing, or browse by family.

Start from the failure

First the pain, then the pattern. Each failure lists the patterns to reach for, closest fit first.

The agent says done, but the work is not actually done.

The workflow has no external signal for completion.

The agent reviews itself and misses obvious problems.

The verifier is too close to the generator.

The same prompt produces different outcomes across runs.

Variance is leaking through sampling, state, timing, or judge behavior.

The check passes because the environment already looked right.

The verifier observes a true fact, but not causality.

Agents disagree, loop, or escalate randomly.

The system has no explicit routing policy for uncertainty.

Tool calls are messy, unsafe, or hard to verify.

The boundary between model intent, tool input, and policy is ambiguous.

Browse by family

Context and State

What exists before the model acts. 5 patterns.

Verification

How to turn judgment into observable signals. 6 patterns.

Orchestration

How to control bias through independence and feedback routing. 6 patterns.