Resources
Examine the assumptions behind your agentic system.
Choose a focused check from the engineering resource library, work through its questions and use the result to decide what evidence or investigation you need next.
Start with the overall design
Agent Design Check helps you identify unanswered questions about purpose, evidence, permissions, actions, evaluation, recovery and ownership. Its checklist shows what still needs attention. It does not assign a production-readiness score.
Scope and control
- POC Selection Tool: examine whether a use case has a credible path for further work.
- Agent Authority Review: examine which workflow steps need code, a person or bounded agent action.
- Rule Placement Audit: record where each important rule is enforced.
- Gate Routing Map: distinguish cases that should block, wait or go to a person.
Evidence and failure
- Citation Fit Check: test whether cited evidence applies to the situation.
- Failure Triage Quiz: practise on an illustrative incident.
- Failure Triage Kit: work through what the system should do after a failure or rejection.
- Agent Memory Audit Kit: examine memory structure and failure cases.
Model choice and cost
- Model Selection Tool: assess a candidate model for a defined step.
- Rework Cost Check: examine retries, reviews and repeated work.
- Cost-Ceiling Workbook: estimate spend and examine stopping boundaries.
- Run-Cost Model: compare ways of completing the same job over one period.
Read the result in context
Each tool states its method and limits. A score or completed checklist supports a discussion; the people responsible for the system still need to examine the evidence and make the operating decision.
Use illustrative or non-confidential information while exploring. Check the resource's data-handling explanation and Privacy before entering information about your system.
Browse written guides · Explore guided build programmes · Discuss a design review