What I Learned Using Free AI Coding Agents on a Real Portfolio (OpenCode and Kilo)
说明与用途
在 VS Code 中使用免费 AI 代码代理加速了个人作品集的开发,但也导致错误和延迟。作者制定了逐步流程:只读诊断、小幅更改、人工审核、本地构建和部署检查,从而降低了错误率。
问题
Relying on free AI coding agents in VS Code caused issues such as incorrect fixes, stalled execution, accidental commits, and environment mismatches.
结果
Adopting a disciplined, step‑by‑step workflow with AI agents reduced errors, prevented accidental commits, and ensured reliable builds and deployments.
工作流
- Ask for a read‑only diagnosis
- Ask for one small change
- Run git status and git diff --stat
- Run the build locally
- Test the page on desktop and phone-sized window
- Commit with a clear message and push
- Check deployment status is Ready
- Verify live site in incognito tab
证据
They saved me a lot of time, but they also got stuck, lied by accident, and once looked busy for a whole hour without doing anything.
What worked was splitting the job into two steps: First, read only. Do not change anything. Find where the mobile menu is rendered and tell me exactly why the panel is invisible, with the file and line number. Then fix it.
Short prompts with one job each worked much better.
Agents are happy to "help" by committing and pushing. I wanted to review every change first.
Typecheck passing does not mean the build passes
- 记录日期: 2026-11-02T10:14:08.252Z