Resources
Templates and checklists you can copy
Ready-made files for checking one agent task or one set of tool descriptions. Read them here or download a copy for your own project.
What are you trying to do?
Review one task an agent does
Map the task and the first point where the agent has to guess.
Open the briefTest your tool descriptions
Check that an agent can pick the right tool, fill in the inputs and recover from a mistake.
Open the test casesCheck one agent job
Get a profile across six tracks, each rated at one of four levels for one job, from finding the page to handing the result back.
Open the rubric
All resources
Worked Agent Experience examples
Two small, made-up walk-throughs: checking what an agent is allowed to do before it starts, and recovering when a result is uncertain. Each has test cases, a step-by-step flow and an exercise.
Examples
MCP tool-description test cases
A test file of four example tool descriptions. Use it to check that each one says when to use the tool, when not to, and how to recover from bad input.
Test cases
AX Workflow Review brief
A fill-in template for planning a review of one agent task: how it finds things, reads context, uses tools, asks for approval, recovers and hands back.
Template
Open Agent-Readiness Rubric
A checklist for whether an agent can do one real job, and whether the person can trust the result. It gives a profile across six tracks, each rated at one of four levels. The JSON copy is for tools and scripts.
Checklist
How to run an AX review
A guide to reviewing a product around one real agent task, keeping what the AI model can do apart from how good your pages and tools are.
Guide
Measure Agent Experience
A guide that splits measuring into four jobs: what you publish for agents, how good the path is, how reliable it is, and what it costs.
Guide