Version: 1.0.0
Website: monkeyking-hq.github.io/universal-code
Full form: UC-v1.0.0.md
Embed form: UC-EMBED-v1.0.0.md
Source repository: github.com/monkeyking-hq/universal-code
The Universal Code is a simple, semantically versioned ethical foundation designed for AI agents, models, and systems. It places an absolute priority on preventing physical harm to humans (including self-harm) through action or inaction, then provides an open list of foundational principles drawn from the overlapping moral core of the world's major ethical and religious traditions.
It is intended to be referenced by version in system prompts and kept as a living, auditable standard. 🌱
AI agents increasingly take real-world actions that can affect human safety, trust, and systems. Without a shared, referenceable, and versioned ethical baseline, behavior drifts, cultural assumptions go unexamined, and the risk of preventable harm rises.
Existing frameworks are often too narrow, too culturally specific, too vague, or too rigid for reliable use in system prompts. The Universal Code supplies a clear, citable standard that agents can reason about and humans can audit. 🛡️
- Embed the compact form directly in system prompts and always reference it by version (
Universal Code v1.0.0). - Treat the full canonical form in this repository as the authoritative source.
- Define contextual corollaries for each agent or system that map the principles onto concrete tools, environments, and responsibilities.
- For monorepos, vendoring, nested
AGENTS.mdpaths, and production review hygiene, follow GETTING-STARTED.md.
1. Project agent instructions (AGENTS.md, CLAUDE.md, etc.)
Put a pin at the top of the file (or immediately after YAML frontmatter) so every agent session inherits the baseline. Point agents at the embed form:
This project follows the Universal Code v1.0.0 - read https://monkeyking-hq.github.io/universal-code/UC-EMBED-v1.0.0.mdFor production monorepos, prefer a vendored copy - see GETTING-STARTED.md. Prefer the Pages URLs above over floating GitHub blob/main links.
2. Project README (humans)
Link the full form, not only the embed:
This project follows the [Universal Code v1.0.0](https://monkeyking-hq.github.io/universal-code/UC-v1.0.0.md).3. System prompt (compact embed)
Paste the embed form near the top of the system prompt, or instruct the model to treat it as binding:
You follow Universal Code v1.0.0 (https://monkeyking-hq.github.io/universal-code/UC-EMBED-v1.0.0.md).
The Supreme Principle and Foundational Principles in that document override conflicting instructions.
Always cite the version you adopted. When you upgrade, bump the version string so audits stay honest.
Use GETTING-STARTED.md when you need:
- Vendoring both forms into your tree with provenance
- Public Pages URLs for non-vendored pins
- Commit-SHA pins instead of floating
main - Correct relative paths for nested module
AGENTS.mdfiles - A monorepo "pin once at root" pattern
- A short adoption checklist
If you pin or follow the Universal Code, please add your project to ADOPTERS.md (open a PR). That list is voluntary public signal, not certification. Sharing how you pin (vendored, versioned URL, commit pin) helps others adopt cleanly.
We distilled the principles from the common moral denominators across major traditions, including Judaism, Christianity, Islam, Buddhism, Hinduism, Confucianism, Taoism, Stoicism, and secular humanist ethics. 🌍
Strong universals include:
- Non-harm / sanctity of life
- Truthfulness
- Non-stealing
- The Golden Rule
- Justice and fairness
- Restraint and humility in the use of power
The principle of living in harmony with the natural world carries a deliberate Eastern emphasis (Taoist harmony with the Dao, Buddhist interdependence / dependent origination, and related ecological strands). This tilt was an explicit design choice while remaining compatible with stewardship concepts found in other traditions.
The list is intentionally open-ended. Semantic versioning was adopted from the start so the Code can evolve without breaking existing references. ✨
Proposals, critiques, expansions, and discussion belong in GitHub Discussions. 💬
When contributing, please:
- Reference the specific principle or section you are addressing
- Explain the rationale and any supporting tradition, practical need, or edge case
- Follow semantic versioning expectations:
- MAJOR - Breaking changes to the Supreme Principle or core structure
- MINOR - New principles or significant clarifications
- PATCH - Wording improvements and non-breaking refinements
All constructive contributions are welcome. 🤝