Example prompts
A request names what to do and, when it runs something, where to run it. A machine is named by host name, by address or by operating system, as it appears in the endpoints file, or left to the run as any available endpoint that fits. The prompts below are examples; any request about the suites, the machines and the logs works the same way.
Create a suite
The source is the documentation of the system under test: a README, a user manual, an API specification or the code itself. The AI agent creates the suite repository laid out like the tutorial suite.
Create a suite for the maze application from its README.
Run
The AI agent fills in the endpoints and the values in a file of its own, runs it and reports from the log. The test itself is never edited.
Run the TCP Traffic group: Linux machines as fan-out, Windows machines as link partners.
Run the network suite with the macOS machine as SUT and lab-win-1 as link partner.
Run TCP Transmit on any available Linux endpoint.
Add a test
The AI agent writes the new component in the suite repository, in the structure of the existing ones, and can run it right away on the machines named.
Add a bidirectional traffic test to the TCP Traffic group, fail below 900 Mbit/s, and run it on lab-linux-1.
Analyze results
The AI agent reads the logs of earlier runs and answers from what they record.
Analyze the results of the nightly validation run.
Explain why TCP Receive failed on lab-linux-2 in the last run.
Scheduled prompts
A request given once and run on a schedule keeps the suite growing with the system under test. Each pass ends with the new tests run once and a commit for you to review.
Every Friday, compare the maze application's features with the maze suite and add tests for what is not covered.
Every night, for every bug closed today in the maze tracker, add a test to the maze suite that reproduces it.