AI-Powered Test Agent (Proof of Concept)

This repository contains the proof-of-concept for an AI agent and Gemini CLI extension that automatically generates test cases based on code review feedback.

It includes:

1. A sample Node.js application (/src).
2. A Gemini CLI extension (/test-agent) with custom commands.
3. A pre-staged test scenario (/review-scenario) to demonstrate the workflow.
4. The PRD and Design documents for the project.

Installation and Setup

To get started, you will need node, npm, and the gemini CLI installed.

1. Clone the repository:

   git clone <your-repo-url>
   cd test-agent

2. Install project dependencies: This installs jest for the sample application.

   npm install

3. Link the Gemini CLI extension: This command makes the /generate-test-from-finding and /commit-test commands available to the Gemini CLI. It must be run from the project root directory (test-agent/).

   gemini extensions link test-agent

How to Test the Workflow

After completing the installation steps above, you can test the full workflow in the pre-staged scenario.

1. Navigate into the scenario directory:

   cd review-scenario

2. Run Code Review: Ask the agent to review the staged changes.

   gemini /code-review

   You should see a report identifying a division-by-zero risk in src/calculator.js.

3. Generate a Test for the Finding: Use our custom command to generate a test for the finding from the report.

   gemini /generate-test-from-finding src/calculator.js 2

   The agent will suggest a new test case and instruct you to run /commit-test.

4. Commit the New Test: Run the commit command to apply the suggested test to your local file.

   gemini /commit-test

   (Note: In this MVP, this step will fail due to CLI sandboxing, but it demonstrates the intended workflow.)

5. Verify the Test Fails: Manually apply the suggested test to src/calculator.test.js and run the test suite. The new test should fail, proving it caught the bug.

   npm test