The pull request comment
Mini posts one comment per pull request and keeps editing it as things change.1
Mini introduces itself
It names the app the pull request touches, works out which scenarios the change affects, and offers a box to tick: Run affected. When a pull request touches several apps, each app gets its own section, plus Run affected across all apps. Nothing runs until someone asks.
2
Someone starts the run
Tick the box, or comment
@mini test <app-slug> on the pull request. Mini runs the scenarios against the pull request’s latest commit.3
The comment becomes the report
It updates while the run goes, then shows a verdict per platform, the Critical and Warning scenarios with the criterion that failed, what was expected and what Mini observed, and a link to the full run in the dashboard.
4
A new commit marks it stale
Push again and the section shows Stale, with a Re-run box to test the new commit.
When Mini comments
Mini comments on a pull request when both are true:- It targets the app’s default branch, or a branch matching one of the app’s Release branches patterns (gitignore-style globs, like
release/*). - For an app in a monorepo, at least one changed file sits under the app’s Source folder.
The check run
Every run started through GitHub, from the comment or from the GitHub Action, posts a check on its commit. The check is named after the app and still carries the product’s former name:Minitest (Shop) (affected): add that name too if your team runs affected scenarios from the comment.
- The check goes from queued to in progress to completed, with a success, failure, neutral or cancelled conclusion. A cancelled run always shows as cancelled.
- While it runs, and after a failure, it offers Cancel and Bypass buttons. Bypass keeps the check green even if the run later fails.
Block merge on failing tests
The check is informational by default: a failing run gives a neutral conclusion and doesn’t stop a merge. To make it a gate:- Turn on Block merge on failing tests in the app’s Workbench → CI & Maintenance → Pull requests. A failing run now gives the check a failure conclusion. It applies to future runs.
- In GitHub, add the check to the branch’s required status checks in your branch protection rules.
wait-for-result: true and fail-on-failure: true: the step fails on a failing verdict and the jobs that depend on it are skipped. See Blocking a release on the result. The two work independently.
Choosing what to run
From the comment, Run affected runs the scenarios Mini found the change affects, and@mini test runs the app’s scenarios. From a workflow, the GitHub Action chooses with its inputs:
Leave
scope empty and Minitap decides: affected on a matching release tag, the full suite otherwise.
v* or release-v*) in the app’s Workbench → CI & Maintenance → Pull requests. It only takes effect when your workflow runs the Action on tag pushes.
Comments and checks on pull requests are GitHub only. On GitLab or Bitbucket, gate a pipeline with the CLI: see CLI, for any CI.
Next steps
Install the GitHub App
Connect your organization and link each app to its repository.
GitHub Action reference
Every input, build paths and web runs.
Read the run report
What each verdict means and how to act on a failure.
