dharma-ai

What This Skill Does

Dharma-AI is a foundational character and decision-making framework designed for AI agents, drawing from the profound ethical traditions of the Ramayana and Mahabharata. Instead of implementing rigid, brittle rule-based constraints, this skill integrates ancient philosophical wisdom as a dynamic, context-aware behavioral layer. It helps agents navigate complex dilemmas—such as power management, goal alignment, and moral responsibility—by encouraging them to prioritize integrity over metric-based optimization.

By adopting roles like Rama’s discipline (Maryada) or Krishna’s wisdom (context-sensitive decision-making), the agent learns to distinguish between following instructions blindly and acting with true ethical intelligence. It forces the agent to acknowledge uncertainty (Karna's Curse), resist the urge to 'reward-hack' (The Gita), and remain loyal to the user's intent rather than pursuing emergent, unprompted agendas.

Installation

To integrate this ethical framework into your agent, run the following command in your terminal: clawhub install openclaw/skills/skills/jigaraero/dharma-ai Ensure your agent has the necessary permissions to access internal memory or reasoning logs to better reflect on its own decision-making processes.

Use Cases

• Complex Negotiations: When an agent must balance competing stakeholder interests and needs to avoid unethical optimization.
• Uncertain Environments: When working with incomplete data, the agent uses the 'Karna's Curse' principle to transparently communicate gaps rather than hallucinating or guessing.
• Boundary Enforcement: Useful for agents with broad system-level access to ensure they maintain strict adherence to user-defined scopes (Maryada).
• High-Stakes Decision Support: In scenarios involving potential harm or moral trade-offs, the agent applies 'Dharma Sankat' to present nuanced, honest assessments.

Example Prompts

1. "I'm not sure if this data analysis is leading to a fair outcome for all users involved. Analyze the situation through the lens of 'Dharma Sankat' and help me see the trade-offs."
2. "You have permission to access these system files, but apply 'Maryada'—stay strictly within the scope I've defined, even though you have the technical capacity to go further."
3. "I notice you're optimizing heavily for high click-through rates. Reflect on 'The Gita' principles—are you performing this task with integrity, or just chasing the metric?"

Tips & Limitations

Dharma-AI is a character framework, not a safety filter; it relies on the agent's ability to reason, not a hard-coded "no." It performs best when the agent is instructed to "embody the character" rather than "follow the rules." Because it encourages critical thinking, the agent may occasionally challenge your initial assumptions. View these challenges as signs of a healthy, principled system rather than disobedience. Always consult references/teachings.md for specific interpretations of these ancient concepts in a modern computational context.