Clacky AI Best Practices
Never Submit Business Requirements in the Root Thread
Root Thread should only be used for setting up the development environment when first importing a project, committing .1024 to the code repository, and making necessary updates to .gitignore. Never submit business requirements in this environment.
This is just where you first experience the Clacky workflow. In the future, you should only update necessary development environment dependencies, and it will automatically stay synchronized with the code in the development main branch.
Don’t Try to Make Clacky Build Projects from 0 to 1
Clacky is designed to help programmers handle iterations of existing historical projects, whether it’s fixing bugs, adding requirements, or maintaining technical debt, it can handle it well.
For completely new projects, we encourage you to use a framework scaffold, such as rails new new_project to generate a sample project, then upload it to GitHub or GitLab, and then import it into Clacky to start.
You can also generate projects from bolt/lovable, upload them to GitHub/GitLab, and then import them into Clacky for development.
Each Thread Should Handle Only One Clear Requirement
Each Thread is an independent working branch and an independent development environment. With clean context, it can better leverage ClackyAI’s coding capabilities. Multiple unrelated requirements may challenge ClackyAI’s attention.
Encouraged: Try to Break Down into Medium-Scale Requirements
Requirements that would take junior to mid-level engineers about 3-5 days of work are more suitable for ClackyAI to complete. If your requirement is larger (more than 20 steps), it’s recommended to break it down using your professional judgment and iterate the requirements in multiple sessions.
Encouraged: Sometimes You Don’t Need Detailed Requirement Descriptions, Fully Utilize AI’s Creativity
One-sentence requirements are still encouraged.
Encouraged: Use Programmer Language to Present Clearer Requirements
Good:
I found the issue might be in configuration, causing xxx feature not to work properly, please try
Bad:
xxx feature doesn't work, please fix it
Known Requirements That Are Not Well Handled (We Are Improving)
- Installing complex dependencies (manual installation and handling is recommended), we have preset a batch of middleware (mysql, redis, mongo, postgres, etc.) recommended for use
- Using libraries with external dependencies that have undergone multiple major API iterations, AI will have confused memories