Skip to main content
When a task completes, it enters Testing status for manual review. With Auto Test enabled, an AI agent automatically verifies the work before you see it.

​
How It Works

  1. Task completes and enters Testing status
  2. A verification agent starts automatically
  3. The agent reviews code changes, runs tests, and checks acceptance criteria
  4. Pass — task moves to Done automatically
  5. Fail — task stays in Testing with failure details logged

​
Enabling Auto Test

Open Project Settings > Automation tab:
  1. Toggle Auto Test to enabled
  2. Optionally add Custom test instructions for project-specific checks

​
Custom Instructions

Add project-specific test commands in the text area: 
Run 'npm test' and verify all tests pass.
Check TypeScript compilation with 'npx tsc --noEmit'.
Ensure no console.log statements remain in production code.
Verify that new API endpoints have corresponding test files.

​
What Gets Verified

The verification agent automatically:
  • Reviews the git diff of changes made
  • Runs the project’s existing test suite
  • Checks each acceptance criterion (if specified on the task)
  • Performs a build/compile check
  • Looks for common issues (broken imports, missing error handling)

​
Verdict Outcomes

VerdictActionDescription
ApproveTask → DoneAll checks passed
RejectStays in TestingIssues found, feedback logged
UnknownStays in TestingCould not determine result, manual review needed

​
Activity Tracking

All auto-test actions appear in the activity timeline:
  • test_started — Verification began
  • test_passed — All checks passed, task approved
  • test_failed — Issues found, details in task logs
Auto Test uses the Sonnet model with low thinking effort for cost efficiency. Custom instructions override the default verification behavior.