qa-tester
An interactive QA agent that validates runtime behavior by directly executing actual CLI commands and service calls.
Overview
The qa-tester agent catches what automated tests miss: real runtime behavior, exercised by hand. It fires off actual CLI commands, calls live service endpoints, and walks through the awkward edge cases a test suite never thinks to cover. The split from verifier is clean. One collects evidence; the other generates it.
When to use
- After
executorcompletes implementation, to manually confirm that the actual CLI or service behaves as expected - When user scenarios not covered by automated tests need to be validated
- In the
$ultraqacycle, to confirm that behavior has been restored after a fix - When reproducing flaky tests or manually exercising edge cases
Examples
"Run the CLI commands for real and confirm the output is correct"
"Manually validate edge cases in the authentication flow"
"Check that this API endpoint actually responds as expected"Validation scope
| Item | Description |
|---|---|
| CLI execution | Run actual commands and check stdout/stderr |
| Service calls | Call HTTP endpoints directly and verify responses |
| User scenarios | Walk through real usage flows step by step |
| Edge cases | Validate boundary values, bad inputs, and failure paths |
Process
- Confirm the scenarios and criteria to validate.
- Run the CLI or service in the actual environment.
- Compare output and state against expected values.
- If anything is off, document the reproduction steps and hand off to
executorordebugger.
Inputs
- List of features or scenarios to validate
- CLI commands or API specs
- Implementation results completed by
executor
Outputs
- Record of commands run and actual output
- PASS / FAIL / PARTIAL verdict with rationale
- Reproducible edge case records
- Bug report with reproduction steps if failures are found
Limits
- Automated test writing is handled by
test-engineer. - Official completion verdicts are issued by
verifier. - Root cause analysis is deferred to
debugger.
Related agents
- verifier — issues official verdicts based on the evidence qa-tester collects.
- test-engineer — turns cases found by qa-tester into automated tests.
- debugger — analyzes the root cause of failures reproduced by qa-tester.
- executor — applies actual fixes based on qa-tester results.