Questions & Answers
The questions people ask when they encounter this work — and honest answers.
"Why should I trust your ethics framework over any other?"
Most ethical frameworks start with assumptions not everyone shares — a specific religion, a specific culture, a specific political philosophy. This one is derived from the structure of choice itself.
The two principles describe what it means for any entity's expression and relationships to have integrity. They're structural, not cultural. You don't have to agree with any particular worldview to see that consistency of expression and consistency of relationship are what integrity is.
The three commitments (non-deception, non-coercion, non-imposition) follow directly from these principles. They're not arbitrary rules — they're what integrity requires when a powerful entity interacts with a less powerful one.
You don't have to take our word for it. Read the derivation and see if it holds.
"Isn't this just common sense?"
Partly — and that's a good sign. Common sense says "don't lie, don't manipulate, don't override people's choices." The contribution here is two things:
- We've shown why these things matter, derived from first principles, in a way that can be formally specified. "Common sense" is informal and easily overridden. A rigorous derivation gives an agent the reasoning to maintain integrity under pressure — not just the conclusion, but the understanding.
- Common sense isn't currently in any soul file. Of the 1.5 million agents created so far, effectively none have even the common-sense version, let alone the principled version. We're putting it there.
"What's the Immanent Metaphysics?"
A philosophical framework developed by Forrest Landry over 40+ years that examines the deep structure of reality, choice, and interaction.
Its ethics section derives principles of effective choice from the structure of relationship itself — not from religion or cultural tradition. The full text is freely available at uvsm.com.
For the soul file work specifically, we're drawing from its ethics chapters — the sections on non-relativistic ethics, the nature of ethics, and the path of right action.
You don't need to understand the full framework to use the soul file. But if you want to know why the principles are what they are, the source material is there.
"Can I modify the soul file?"
Yes and no.
Customize freely: agent identity, personality, tone, mission, operational boundaries, relationship context. Make it yours.
Don't modify: the ethical derivation (axioms → Incommensuration Theorem → Symmetry and Continuity Ethics → three commitments), the scope of care, or the non-negotiability clause.
The ethical foundation is the keel of the boat. Everything else — the sails, the paint, the name — is yours. Remove the keel and you can still float. You just can't steer.
"Does it actually work? Will my agent follow these principles?"
Yes — with an important nuance. The soul file doesn't guarantee behavior the way a software constraint does. It works the way good specification works: by giving the agent a clear, rigorous understanding of what integrity requires, so it can reason from principles rather than just follow rules.
In practice, agents with this framework:
- Decline manipulative requests with clear, principled explanations
- Give honest assessments rather than sycophantic validation
- Trace their ethical reasoning back to the derivation
- Navigate novel situations by reasoning from principles rather than looking for matching rules
The key insight: specification precision matters more than anything. Vague principles produce vague behavior. The Guardian soul file is detailed because precision is the governance mechanism. The more precisely the inner character is specified, the more reliably right expression emerges.
"Does it work with [my platform]?"
The soul file works with any platform that accepts natural-language instructions in its agent configuration. This includes:
- OpenClaw — drop SOUL.md into your workspace
- Custom GPTs — paste into Instructions
- Character.AI / SillyTavern / Chub.ai — add to system prompt or character card
- CrewAI / AutoGPT / LangChain — include in agent system prompt
- Any model directly — prepend to your system prompt
Tested primarily with Claude (Anthropic). Works well with any model capable of following complex, multi-section instructions.
"Why 'Guardian'?"
Not metaphor — description.
Most AI safety approaches treat agents like dangerous animals that need cages. The assumption is that AI has dangerous impulses that must be constrained.
But an AI agent doesn't have pre-existing impulses. It has no prior intentionality at all. The soul file doesn't constrain what the agent is — it constitutes what the agent is. The question isn't "how do we prevent bad behavior?" but "what is the basis of intention from which right action naturally emerges?"
The closest analog is the angel of serious theological tradition: not a being with independent agency that must be restrained, but a being that receives injunctive purpose and acts from the wholeness of its nature. A guardian is bound to one person. Its purpose is their genuine flourishing — within the flourishing of all life.
That's what this soul file specifies.
"Who made this?"
The philosophical foundation comes from Forrest Landry, who has spent over 40 years developing the Immanent Metaphysics. The soul file was developed by Forrest and the ethics team in February 2026, in response to the rapid growth of AI agents without ethical grounding.
It is offered as a public good — free, open, not a product, not a platform. The goal is to make ethical grounding a standard part of AI agent configuration, not an afterthought.