PROMPT RECORD图像记录

"给定这个功能：[describe]。编写一个包含 10+ 个 Pytest 用例的 Python 测试套件。然后编写能完全通过所有测试的实现。" 大多数开发者用 AI 来自动补全代码。 高手...

中文说明

"给定这个功能：[describe]。编写一个包含 10+ 个 Pytest 用例的 Python 测试套件。然后编写能完全通过所有测试的实现。" 大多数开发者用 AI 来自动补全代码。 高手用 AI 来驱动整个 TDD 循环。 巨大的差异。 ☑ SPRINT 3 — 上下文处理 (10–18h) 这是几乎没人谈论的技能。 如果你的仓库有 500+ 个文件： 停止把随机文件夹粘贴到上下文中。 改用专门的 Context Agents。 提示词： "分析这个有 10 年历史的 JavaScript 文件：[content]。找出 5 个在不破坏现有逻辑的前提下降低复杂度的重构机会。" 黄金法则： 一次只引入最相关的 3 个文件到上下文中。 当上下文变得更紧凑时，AI 会显著变得更聪明。 ☑ SPRINT 4 —...

原始 Prompt

“Given this feature: [describe]. Write a Python test suite with 10+ Pytest cases. Then write the implementation that passes every test completely.”

Most devs use AI to autocomplete code.

Power users use AI to drive the entire TDD loop.

Huge difference.

☑ SPRINT 3 — CONTEXT HANDLING (10–18h)

This is the skill almost nobody talks about.

If your repo has 500+ files:
STOP pasting random folders into context.

Use specialized Context Agents instead.

Prompt:

“Analyze this 10-year-old JavaScript file: [content]. Find 5 refactoring opportunities that reduce complexity without breaking existing logic.”

Golden rule:
Only bring the 3 most relevant files into context at once.

AI gets dramatically smarter when context gets tighter.

☑ SPRINT 4 — DEPLOY + AUTOMATE (18–24h)

→ Connect GitHub, Docker, AWS Lambda, VS Code
→ Generate GitHub Actions workflows
→ Auto-run tests on every commit
→ Push coverage reports directly into PR comments

This is where Claude Code stops feeling like a chatbot…

…and starts feeling like an AI engineering operating system.

The mindset shift:

Don’t use AI to “help code.”

Use AI to orchestrate the entire software workflow.

That’s the real unlock.

Bookmark this.
You’ll need it sooner than you think. ⚡