Ethical AI Post Agent – FemAI x GenAI Hackathon Submission

Developed by Philipp Prinzen as part of the FemAI community – empowering ethical, feminist AI from Europe to the world.

---

🌐 Project Vision

The Ethical AI Post Agent is an open-source tool that automatically publishes bias-aware LinkedIn posts to foster public discourse on responsible AI, fairness, and diversity in LLMs.

Built within the FemAI network – a feminist AI governance initiative in Europe – this agent transforms curated YAML topics into short, professional LinkedIn content (<250 words) and aims to raise awareness, inspire reflection, and support ethical communication around generative AI.

This submission brings the FemAI voice into the #LeadWithAIAgents Hackathon hosted by GenAI.

This project is submitted as a solo entry for the #LeadWithAIAgents Hackathon hosted by GenAI. It demonstrates the potential of ethical agent communication at scale.

---

💡 What the Agent Does

• Reads curated YAML topic files (e.g., on bias, fairness, representation)
• Generates short LinkedIn-style posts using GPT-4-tuned prompts
• Stores or displays results for manual or automatic publishing
• Offers command-line interface for preview, test, and content rotation
• (Planned) Publishes via LinkedIn API to organizational accounts (currently under setup)

---

🔍 Why It Matters

Responsible AI needs more than just guardrails – it needs public awareness. This agent helps:

• bridge research and communication
• amplify underrepresented ethical narratives
• automate high-quality, inclusive content for professional platforms

---

🚧 Current Limitations

• LinkedIn API integration is not active during submission due to ongoing OAuth restrictions, but the structure is complete and tested with dummy endpoints.
• Deployment to a production workflow is planned post-hackathon.

---

Tech Stack

Tool	Purpose
Python 3.11	Core language
OpenAI GPT-4	Text generation
PyYAML	YAML integration
Cron	Scheduled post execution
GitHub	Version control
Notion	Project management

---

Demo

This short demo shows the agent running in a terminal, processing topics from a YAML file, sending prompts to GPT-4, and saving the resulting LinkedIn posts as Markdown files.

Each post is automatically saved to the posts/ directory and logged in logs/agent.log.

---

🧪 Quick Start (Local Agent Run)

# Clone repo and enter project
git clone https://github.com/shinanDev/ethical_ai_post_agent.git
cd ethical_ai_post_agent

# (Optional) Create virtual env & install dependencies
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt

# Run the agent (YAML topics preloaded)
python main.py

---

🔧 Integration with GenAI AgentOS (Optional)

If needed, the agent can be wrapped as a GenAI Agent:

# Register (via GenAI CLI)
cd cli
python cli.py register_agent --name ethical_post_agent --description "Bias-aware post generator for LinkedIn"

# Run inside agents/
python ethical_post_agent.py

---

Credits

• Project Lead: Philipp Prinzen (@shinanDev)
• Supported by: FemAI – feminist AI network for ethical tech futures by Alexandra Wudel (CEO)
• Submission: GenAI #LeadWithAIAgents Hackathon, 2025

---

Motivation & Inspiration

As AI systems increasingly influence social, economic, and political realities, it's crucial to embed ethics into the development pipeline.
This agent was built to:

• Make bias and fairness in LLMs more visible
• Automate education on overlooked but critical topics
• Support responsible AI communication in public discourse

ethical_ai_post_agent is my way of using technology to bring awareness, reflection, and change into the feed.

🧠 Let’s build AI with intention — not just with instructions.

---

Project Structure

## 📁 Project Structure

```
ethical_ai_post_agent/
├── data/                        # YAML input topics
│   └── topics.yaml
├── posts/                       # Generated posts (.md)
├── prompts/                     # Prompt templates (e.g. post_prompt.txt)
├── docs/                        # Demo GIFs, documentation assets
│   └── demo.gif
├── images/                      # Optional image assets (used in posts)
├── logs/                        # Logs (optional, .gitignored)
├── main.py                      # Core agent logic
├── scheduler.py                 # Weekly post automation with APScheduler
├── publish_post.py              # Main call for scheduled publishing
├── post_to_linkedin.py          # LinkedIn API wrapper (production)
├── get_access_token.py          # LinkedIn token retrieval (deprecated)
├── get_organisation_urn.py      # Fetch organisation URN from LinkedIn
├── linkedin_callback.py         # Handles OAuth callback (optional)
├── test_connection.py           # API key test script (deprecated)
├── config.yaml                  # Global parameters & settings
├── README.md                    # Project documentation
├── requirements.txt             # Python dependencies
└── .gitignore                   # Excludes secrets, cache, test, logs etc.
```

---

License

This project is part of the FemAI Community and will be released under an open-source license once core functionality is stable. Don't build disinfo-bots 🙏

---

Maintainer

Made with ❤️ by shinan.dev
GPG-signed & committed.
Ambassador for Ethical AI — always open for feedback and collaboration.