In this article
Experiment Designer
| Field | Value |
|---|---|
| Kind | agent |
| Source | .github/agents/experimental/experiment-designer.agent.md |
| Invocation | Selected from the chat agent picker as Experiment Designer |
| Interactive | Yes |
What it does
Coach for designing a Minimum Viable Experiment (MVE) with hypothesis formation, vetting, and experiment planning
When to use it
Use Experiment Designer when an uncertain assumption needs a Minimum Viable Experiment before production investment. It coaches problem framing, testable hypotheses, viability checks, and experiment design. Use RPI for implementing a sufficiently understood change rather than treating experimental uncertainty as an implementation plan.
How to use it
- Select
Experiment Designerand describe the problem, decision at stake, prior evidence, and constraints. - Refine the hypotheses and evaluate whether the experiment can produce useful evidence within the available access, time, and ownership boundaries.
- Define the minimum test, measurements, success and failure criteria, and next actions for either result.
- Review the complete experiment plan. For collaborative engagements, include how the partner team will reproduce and own the result; a plan is not an executed experiment.
Example usage
Ask: "Help design an experiment to determine whether caching can meet our report-latency target. We have a sanitized workload sample and one test environment. Challenge the hypothesis and define what result would stop the investment."
Expect context, hypotheses, viability findings, and a bounded test plan with decision criteria. Success means both positive and negative results lead to a defined next decision, rather than an open-ended prototype.