Every PR gets
a litmus test.
Dip a grounded persona into your preview deploy. Three minutes later the reaction is on the PR: “I couldn’t find the button.” One comment. Edits itself. Zero noise.
Most UX bugs aren’t caught by tests. They’re caught by users — after launch.
Pixel diffs catch what you can see. Unit tests catch what you can compile. The moment a real user opens your preview and gets stuck, that lives in nobody’s CI. So it ships. So you find out on Tuesday, from a Slack thread.
Dip. React. Read.
Open a pull request.
Litmusly listens for your preview deploy. Nothing to configure beyond the install. The reagent dips into your real rendered preview, not your code, not a mock.
The reagent walks it.
Sam has 90 seconds and a skeptic's patience. He tries your primary CTA the way a Hacker News visitor would. Stagehand drives; the model reads the reaction.
One comment. Edits itself.
Findings tagged blocker, friction, or nit, each with quoted evidence. Force-push and the comment updates in place. Many PRs deserve zero findings, and get zero.
Not a replacement. A third lane.
Seven reagents. Seven ways your PR reacts.
Each reagent declares what it watches for, and what it’s blind to. That’s what keeps the comment quiet.
Ship the version
that tests clean.
Install in two clicks. Open a PR. Read your first reaction in under five minutes.