Reverse Engineering API

Reverse Engineering API Skill

This skill enables you to reverse engineer web APIs by:

1. Controlling a browser with HAR recording enabled
2. Analyzing captured network traffic
3. Generating production-ready Python API clients

Prerequisites

• Playwright MCP: You must have access to Playwright MCP tools for browser control
• HAR Recording: The browser must be configured to record HAR files
• Python: For running analysis scripts and generated clients

Workflow Overview

[User Task] -> [Browser Capture] -> [HAR Analysis] -> [API Client Generation] -> [Testing & Refinement]

Phase 1: Browser Capture with HAR Recording

Starting the Browser

When starting a browser session for API capture:

1. Launch browser with HAR recording enabled via Playwright MCP
2. Generate a unique run ID: {run_id}
3. Configure HAR output path: ~/.reverse-api/runs/har/{run_id}/recording.har

During Capture

Navigate autonomously to trigger the API calls needed:

• Login flows (capture authentication)
• Data fetching (capture GET endpoints)
• Form submissions (capture POST/PUT endpoints)
• Pagination (capture query parameter patterns)

On Browser Close

When the browser closes, note the HAR file location:

HAR file saved to: ~/.reverse-api/runs/har/{run_id}/recording.har

Phase 2: HAR Analysis

Reading the HAR File

HAR files are JSON with this structure:

{
  "log": {
    "entries": [
      {
        "request": {
          "method": "GET|POST|PUT|DELETE",
          "url": "https://api.example.com/endpoint",
          "headers": [...],
          "postData": {...}
        },
        "response": {
          "status": 200,
          "headers": [...],
          "content": {...}
        }
      }
    ]
  }
}

Filtering Relevant Entries

Filter out noise by excluding:

• Static assets: .js, .css, .png, .jpg, .svg, .woff, .ico
• Analytics: google-analytics, segment, mixpanel, hotjar
• Ads: doubleclick, adsense, facebook.com/tr
• CDN resources: cloudflare, cdn., static.

Focus on:

• API endpoints: /api/, /v1/, /v2/, /graphql
• XHR/Fetch requests with JSON responses
• Requests with authentication headers

Extracting Patterns

For each relevant endpoint, extract:

1. URL Pattern: Base URL, path, query parameters
2. Method: GET, POST, PUT, DELETE, PATCH
3. Headers:
   • Required headers (Authorization, Content-Type, custom headers)
   • Optional headers (User-Agent, Accept)
4. Request Body: JSON schema, form data structure
5. Response Schema: JSON structure, status codes
6. Authentication: See references/AUTH_PATTERNS.md

Phase 3: API Client Generation

Code Structure

Generate a Python module with:

{output_dir}/
  api_client.py    # Main API client class
  README.md        # Usage documentation

api_client.py Template

"""
Auto-generated API client for {domain}
Generated from HAR capture on {date}
"""

import requests
from typing import Optional, Dict, Any, List
import logging

logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)


class {ClassName}Client:
    """API client for {domain}."""
    
    def __init__(
        self,
        base_url: str = "{base_url}",
        session: Optional[requests.Session] = None,
    ):
        self.base_url = base_url.rstrip("/")
        self.session = session or requests.Session()
        self._setup_session()
    
    def _setup_session(self):
        """Configure session with default headers."""
        self.session.headers.update({
            "User-Agent": "Mozilla/5.0 (compatible)",
            "Accept": "application/json",
            # Add other required headers
        })
    
    def _request(
        self,
        method: str,
        endpoint: str,
        **kwargs,
    ) -> requests.Response:
        """Make an HTTP request with error handling."""
        url = f"{self.base_url}{endpoint}"
        try:
            response = self.session.request(method, url, **kwargs)
            response.raise_for_status()
            return response
        except requests.exceptions.RequestException as e:
            logger.error(f"Request failed: {e}")
            raise
    
    # Generated endpoint methods go here
    def get_example(self, param: str) -> Dict[str, Any]:
        """
        Fetch example data.
        
        Args:
            param: Description of parameter
            
        Returns:
            JSON response data
        """
        response = self._request("GET", f"/api/example/{param}")
        return response.json()


# Example usage
if __name__ == "__main__":
    client = {ClassName}Client()
    # Example calls

Code Quality Requirements

All generated code must include:

1. Type hints for all parameters and return values
2. Docstrings for all public methods
3. Error handling with try-except blocks
4. Logging for debugging
5. Session management for connection reuse
6. Authentication handling based on detected patterns

Phase 4: Testing & Refinement

Testing the Generated Client

After generating the client:

1. Run the example usage section
2. Verify responses match expected structure
3. Handle any errors encountered

Iteration Protocol

You have up to 5 attempts to fix issues:

Attempt 1: Initial implementation
  - What was tried
  - What failed (if anything)
  - What was changed

Attempt 2: Refinement
  ...

Common Issues

Domain Discovery (Optional)

Before capture, you may want to map the domain to understand its structure.

Using the Mapper Script

Run scripts/mapper.py to quickly discover:

• All pages on the domain or subdomains
• Subdomains

It is useful for generalizing your scripts on multitenants websites.

For example, for Ashby ATS or Workday it's useful to find other companies using this ATS when trying to generalize your script.

python scripts/mapper.py https://example.com

Using the Sitemap Parser

Run scripts/sitemap.py to extract URLs from sitemaps:

python scripts/sitemap.py https://example.com

Output Locations

• HAR files: ~/.reverse-api/runs/har/{run_id}/
• Generated scripts: ./{task_name}

Example Session

User: "Create an API client for the Apple Jobs website"


1. [Browser Capture]
   Launch browser with HAR recording
   Navigate to jobs.apple.com
   Perform search, browse listings
   Close browser
   HAR saved to: ~/.reverse-api/runs/har/{run_id}/recording.har

   Note: you can monitor browser requests with the Playwright MCP

2. [HAR Analysis]
   Found endpoints:
   - GET /api/role/search?query=...
   - GET /api/role/{id}
   Authentication: None required (public API)

3. [Generate Client]
   Create : {task_name}/api_client.py
   
4. [Test]
   Ran example usage - Success!
   
5. [Summary]
   Generated Apple Jobs API client with:
   - search_roles(query, location, page)
   - get_role(role_id)
   Files: ./{task_name}/