Worked Agent Experience examples
Inspect a capability preflight and an uncertain-result recovery flow. Each includes deterministic test cases, a state flow, and a teaching exercise—not a benchmark or client case study.
Agent ExperienceTemplates and test cases for reviewing agent-facing tools and workflows. Use them on the web or take a copy into your own project.
For source discovery, see Awesome Agent Experience, a public library of papers, specifications, and references.
Inspect a capability preflight and an uncertain-result recovery flow. Each includes deterministic test cases, a state flow, and a teaching exercise—not a benchmark or client case study.
Test whether a public MCP tool catalog tells agents when to use each tool, when not to use it, and how to recover from bad inputs.
Scope one workflow before a review so discovery, retrieval, tool execution, approval, recovery, and handoff are visible.
Inspect declared product surfaces before mistaking a checklist for evidence of successful task completion.
Run a product review around one real agent workflow without confusing model capability with surface quality.
Separate declared surfaces, path quality, reliability, and cost of the path into practical measurement work.
Connect the public MCP server to retrieve the guides and rubric while you work.