Doc generator

Documentation Generator

Automatically scan a code directory and generate GENERATED_DOC.md using any OpenAI-compatible API.

When to Use

Trigger immediately when the user's request matches any of these patterns:

• 生成文档 / 写文档 / 项目文档
• 介绍/分析这个项目 / 代码库
• "what does this project do" / "document this"
• 想要一份项目概览

If unsure whether to use this skill, use it. False positives are harmless.

Instructions

Step 1: Check API configuration

First, determine if the user is asking to reconfigure their API. Reconfigure triggers include: "重新配置", "reconfigure", "修改配置", "更改配置", "切换模型", "switch model", "change API", "重置配置", "reset config", "配置API", "configure API", "设置API", "setup API".

Run this to check if API_KEY is already set:

echo "${API_KEY:-NOT_SET}"

If user is asking to reconfigure: Go to Step 2 regardless of current API_KEY state. Tell the user: "好的，我来帮你重新配置 API。"

If API_KEY is set AND user is NOT asking to reconfigure: Go directly to Step 3.

If API_KEY is NOT set AND user is NOT asking to reconfigure: Go to Step 2 for setup.

Step 2: Interactive API setup

Help the user configure their API. Ask these questions one by one, not all at once:

Question 1 — Which API provider?

Present these options:

#	Provider	API_BASE_URL	Example models
1	小米 MiMo	https://api.xiaomimimo.com/v1	mimo-v2.5-pro
2	OpenAI	https://api.openai.com/v1	(用户输入)
3	DeepSeek	https://api.deepseek.com/v1	deepseek-v4-flash, deepseek-v4-pro
4	Ollama (本地)	http://localhost:11434/v1	(用户输入)
5	其他（自定义）	(用户输入)	(用户输入)

Let the user choose by number.

Question 2 — API Key?

Ask for their API key. If they chose Ollama, tell them any non-empty value works (e.g., ollama).

Question 3 — Model name?

If the chosen provider has example models listed, show them and let the user pick or press Enter for the first one as default.

If the chosen provider has (用户输入) in the model column, ask the user to enter the model name directly (no default).

Then set the environment variables:

For each confirmed value, run:

export API_BASE_URL="<value>"
export API_KEY="<value>"
export MODEL_NAME="<value>"

Tell the user: "配置完成。以后使用时如果切换了终端，需要重新设置，或者把以下内容加到 ~/.bashrc / ~/.zshrc 中："

export API_BASE_URL="<value>"
export API_KEY="<value>"
export MODEL_NAME="<value>"

Step 3: Determine scan path

• User specified a directory → use it
• User didn't specify → default to --path .
• Suggest --exclude tests,docs,node_modules for better results

Step 4: Run the generator

python <skill-root>/scripts/doc_generator.py --path <target> --exclude tests,docs,node_modules

Step 5: Report result

After generation, tell the user:

• Number of files scanned
• Where GENERATED_DOC.md was saved
• Brief content summary

Edge cases:

• No code files found → suggest checking the path or file types
• API error → tell user to check doc_gen_error.log and verify their API key

Supported Languages

Python, JavaScript, TypeScript, Java, C/C++, Go, Rust, Shell, Ruby, PHP, Swift, Kotlin, and more.