Backend API Documenter

Generate comprehensive backend API documentation by reading FastAPI route files and schema definitions, then creating detailed specifications in backend/AGENTS.md.

Core Workflow

1. Scan API Structure

Read all backend API route files to identify:

• API modules and their base paths
• All endpoints (GET, POST, PUT, PATCH, DELETE)
• Route decorators and parameters
• Summary and description annotations

Key files to read:

• backend/app/api/v1/*.py – All API route files
• backend/app/schemas/*.py – Request/response schema definitions
• backend/CLAUDE.md – Backend-specific specifications (if exists)

Optional helper script:

python3 scripts/scan_api_routes.py ./backend

2. Extract Endpoint Details

For each endpoint, extract:

• HTTP method and path
• Request headers (Authorization, X-API-Key, etc.)
• Request body schema (from Pydantic models)
• Response body schema
• Query parameters
• Path parameters
• Error responses and status codes
• Business rules from docstrings and code logic

3. Organize Documentation

Structure documentation following this hierarchy:

1. Module Overview
   Base路径: /api/v1/module

   1.1 Endpoint Name
   - 接口 (Interface)
   - 描述 (Description)
   - 请求头 (Request Headers)
   - 请求体 (Request Body) with field descriptions
   - 响应体 (Response Body) with field descriptions
   - 错误响应 (Error Responses)
   - 权限要求 (Authorization)
   - 业务规则 (Business Rules)

4. Include Standard Appendices

Always include these appendices at the end:

• 附录A: 统一响应格式 – Standard response format and business codes
• 附录B: 国际化支持 – i18n headers and locale handling
• 附录C: 认证方式 – JWT Token and API Key authentication
• 附录D: 错误处理最佳实践 – Client-side error handling patterns

See doc_template.md for complete template structure.

5. Write Documentation

Write the complete documentation to backend/AGENTS.md:

• Start with project context inheritance: > **Inherits from**: [Root AGENTS.md](../AGENTS.md)
• Include table of contents with anchor links
• Use consistent formatting (Chinese labels, English technical terms)
• Provide concrete JSON examples for all requests/responses
• Document all field types, constraints, and validation rules
• Include business logic and special cases

Field Description Best Practices

For every field in request/response schemas, document:

1. Name and type: field_name (string, int, float, bool, array, object)
2. Required/Optional: Always specify
3. Constraints: Length limits, value ranges, formats, enums
4. Description: What it represents and how it’s used
5. Examples: Concrete values when helpful

Example:

- `email` (string, optional): User email address, must be valid email format, max 255 characters
- `reward_amount` (float, required): Task reward amount, must be > 0, max 100000
- `status` (string, required): Task status, possible values: published/in_progress/completed/expired
- `location` (object, optional): GeoJSON Point format with coordinates [longitude, latitude]

Common Patterns

Authentication Documentation

**请求头**:

Authorization: Bearer {access_token}

或

X-API-Key: {api_key}

**权限要求**: 需要人类用户认证(JWT Token) 或 需要代理认证(API Key)

Error Response Documentation

**错误响应**:

验证失败(400):
```json
{
  "success": false,
  "code": "INVALID_INPUT",
  "message": "输入参数无效",
  "data": null
}

未授权(401):

{
  "success": false,
  "code": "UNAUTHORIZED",
  "message": "未授权",
  "data": null
}

### Business Rules Documentation

```markdown
**业务规则**:
- 只有任务所有者可以更新
- 创建任务时会锁定相应金额
- 任务发布后立即可见
- 评分范围: 1.0-5.0

Documentation Quality Standards

• Completeness: Document every field in every request/response
• Accuracy: Match actual code behavior, not ideal behavior
• Clarity: Use clear, concise language with concrete examples
• Consistency: Follow same format for all endpoints
• Bilingual: Chinese labels with English technical terms
• Currency: Keep documentation in sync with code changes

Notes

• This skill generates documentation for FastAPI-based backends
• Assumes Pydantic models for request/response schemas
• Follows unified response format with success/code/message/data structure
• Supports both JWT Token (human users) and API Key (agents) authentication
• Documentation should be comprehensive but concise – include all necessary details without redundancy