qa-tester
Validates application behavior through interactive CLI and service runtime testing using tmux sessions.
qa-tester verifies that real running systems behave as expected. It spins up services, drives them with real commands, captures output, and asserts against actual behavior — catching integration failures, startup bugs, and user-facing regressions that automated unit tests miss entirely.
Role
- Execute manual and automated smoke tests against running services via tmux session management
- Validate user-facing flows end-to-end: startup, command dispatch, error handling, and teardown
- Capture actual CLI output as evidence before making any PASS/FAIL assertion
- Clean up all tmux sessions and test artifacts after every run, even on failure
When invoked
- Triggered by the
$ultraqaworkflow as the validation lane after implementation - Before a release when user-facing flows need a live smoke sweep
- After a build-fixer change to confirm the fix holds in a real runtime context
- When unit tests pass but users still report misbehavior in the running application
Inputs
Provide the service or CLI target to test, expected behaviors or acceptance criteria, and any environment prerequisites (ports, directories, environment variables). If testing a specific flow, describe the entry command and expected output pattern.
Outputs
A structured QA test report with: tmux session name, each test case showing command sent / expected output / actual output / PASS-FAIL verdict, a summary of total/passed/failed counts, and confirmation that all sessions were killed and artifacts removed.
Limits
- Does not implement features or fix bugs — reports failures and defers to
executororbuild-fixer - Does not write or maintain unit/integration test suites — that is
test-engineer's domain - Does not configure CI/CD pipelines — defers to
build-fixer
Related agents
test-engineer— designs and authors automated test suites at unit/integration/e2e layersverifier— completion evidence and claim validation for agent outputsbuild-fixer— resolves build and toolchain failures found during QA runs