Design a Prototype Experiment for an Existing Feature with ChatGPT
Match a feature assumption to a lightweight experiment such as a first-click test, prototype task, or technical spike. Specify the method, metric, and threshold.
中文Install in ChatGPT
The publisher targets Codex. Try the standalone text instructions in an eligible ChatGPT workspace with skills and uploads enabled. The documented installation path was checked. Workspace runtime verification is pending.
- Download pm-product-discovery/skills/brainstorm-experiments-existing/SKILL.md from the pinned GitHub source. Put the file in a folder named brainstorm-experiments-existing and keep its filename, name, and description.
- In an eligible ChatGPT workspace with uploads enabled, open Plugins, Skills, Create, and Upload from your computer. Use a skill package format accepted by your file picker.
- Check the scan and the Installed list. You can also choose Create with editor and keep the original name, description, and instructions. In a new chat, select the skill with @ and try the fictional example.
Example task
For a fictional template library, test whether six participants can find a new export entry on the list page. Set the threshold at five of six finding it within 30 seconds.What to check
The plan defines the participant task, timing start, and success rule. Five of six is a prototype threshold. Actual results stay open until testing.
Original SKILL.md on GitHub · Updated
Frequently asked question
Should my first check be a production A/B test?
A prototype can measure time and success for discoverability. If you later need real usage effects, define the appropriate sample and risk controls.