Create a Test Case Generation & Review Skill and Execute It in Lobster

TestHub provides a default set of prompts for test case generation and review. The author first copies these prompts, noting that alternative prompts can also be used.

1. Build the Skill with Trae

The user asks an AI assistant to create a Skill called generator-testcase that first generates test cases using the "case generation" prompt and then reviews them with the "case review" prompt, finally outputting the final version.

Trae workflow

The tool invokes the built‑in skill-creator utility.

It runs a mkdir command to create a directory named after the Skill (e.g., generator-testcase).

The generated content is saved to .trae/skills/generator-testcase/SKILL.md.

2. Import the Skill into Lobster

The Skill can be imported into the open‑source Lobster platform (also compatible with other AI IDEs such as QClaw or AutoClaw). The author demonstrates the import process with screenshots.

3. Generate Test Cases

After uploading a requirement document or providing a textual description, the user invokes the generator-testcase Skill. The author uses SiliconFlow’s Kimi 2.6 model, noting that a free trial account provides enough credits for multiple generations. The generated test cases are shown in several images.

4. Save Results to Excel

The resulting test cases are exported to an Excel file, and the downloaded file is displayed.

5. Quality Considerations

The author emphasizes that generation quality depends on the AI model, the prompt (or Skill) quality, and the surrounding tooling. When possible, a stronger model should be chosen, and prompts can be iteratively refined during practice.

In summary, improving test case generation quality requires simultaneous attention to the model, prompts, and execution environment rather than relying on a single factor.