> ## Content Index
> Fetch the complete content index at: https://darkwebinformer.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# DorkAgent: LLM-powered agent for automated Google Dorking in bug hunting & pentesting
- URL: https://darkwebinformer.com/dorkagent-llm-powered-agent-for-automated-google-dorking-in-bug-hunting-pentesting/
- Published: 2025-12-11T19:33:07.000Z
- Updated: 2025-12-11T19:33:07.000Z
- Author: Dark Web Informer
- Tags: Tools

---

GitHub: <https://github.com/yee-yore/DorkAgent>

---

🤖 LLM-powered agent for automated Google Dorking in bug hunting & pentesting.

0:00 

/0:06 

1× 

## Usage

[](https://github.com/yee-yore/DorkAgent?tab=readme-ov-file#usage)

1. Git clone

git clone https://github.com/yee-yore/DorkAgent.git  
cd DorkAgent

1. Run DorkAgent

python dorkagent.py

The program will:

- **Auto-install** all required packages on first run
- **Prompt for API keys** and save them to `.env` file automatically

Required API keys:

- `SERPER_API_KEY` \- **Required** (Get from <https://serper.dev/>)
- At least **one** LLM API key:
  - `OPENAI_API_KEY` \- For OpenAI models (GPT-4, GPT-5, etc.)
  - `ANTHROPIC_API_KEY` \- For Claude models
  - `GEMINI_API_KEY` \- For Gemini models

For more description <https://medium.com/@yee-yore/llm-powered-agent-for-automated-google-dorking-dcb14d609dc2>

## Customize

[](https://github.com/yee-yore/DorkAgent?tab=readme-ov-file#customize)

1. The number of google results (`serper_dev_tool.py` inside `site-packages/crewai_tools/tools/serper_dev_tool/`)

class SerperDevTool(BaseTool):  
 ...  
 args\_schema: Type\[BaseModel\] = SerperDevToolSchema  
 base\_url: str = "https://google.serper.dev"  
 n\_results: int = 10 # min: 10, max: 100  
 ...

1. Duration of google search results (`serper_dev_tool.py`)

\# https://serper.dev/playground  
  
def \_make\_api\_request(self, search\_query: str, search\_type: str) -> dict:  
 ...  
 payload = json.dumps({"q": search\_query, "num": self.n\_results, "tbs": "qdr:m"}) # Past week: "qdr:w", Past month: "qdr:m"  
 ...

1. Google dorks (`task()`)

\# Reference https://github.com/TakSec/google-dorks-bug-bounty

1. Agents (`agents()`)

\# https://docs.crewai.com/concepts/agents

## Update Log

[](https://github.com/yee-yore/DorkAgent?tab=readme-ov-file#update-log)

- **2025-12-11**: **DorkAgent v1.4** \- Dynamic model selection via API (choose from available models at runtime), added Pydantic models for structured task output validation, improved code quality with CONVENTIONS.md compliance fixes
- **2025-09-29**: Major code refactoring for improved maintainability (@wjdrud2532 PR #3). Split monolithic 800+ line file into modular architecture (config.py, agents.py, tasks.py, utils.py), fixed dependency auto-installation to work before module imports, added timestamps to report filenames (YYMMDD\_HHMMSS format), enforced code conventions with standardized import order and removed all inline comments, maintained 100% backward compatibility with no breaking changes
- **2025-08-11**: Enhanced security reports with specific information disclosure details, fixed critical notification race condition bug, improved attack vector analysis with actual parameters and payloads, added comprehensive development documentation (CLAUDE.md), restored requirements.txt format
- **2025-05-18**: Modified README.md and banner, Added juicy google dorks, Medium article (<https://medium.com/@yee-yore/llm-powered-agent-for-automated-google-dorking-dcb14d609dc2>)
- **2025-04-17**: Removed tasks(old).py, the version prior to prompt engineering; Deleted Google Dork for finding "Confidential" documents (most results were merely informative); Removed Google Dork targeting login panels; Added settings to help avoid LLM provider rate limits; Integrated Gemini Flash 2.0 (free to use and currently considered the best value LLM); Merged tasks.py and agents.py into dorkagent.py for simplified maintenance
- **2025-04-01**: Added hybrid LLM option (GPT & Claude); Added dork `intitle:"IIS Windows Server"`; Applied prompt engineering to tasks.py; Added default depth consideration for subdomain inputs; Added `requirements.txt` for Windows/MacOS compatibility