The Values Layer Most AI Users Don't Know Exists
Part Two: The Method
Protocol Mods govern how the work gets done. Persona Mods govern who does the work. But neither of those answers the most fundamental question about any system: what does it stand for?
That's the Charter Mod. It's the constitutional layer of your AI system — the set of values, principles, and non-negotiables that govern everything else. It doesn't activate for a specific task or load a specific identity. It runs underneath everything, always.
Most AI users don't know this layer exists. They optimize their prompts, maybe build a persona or two, and wonder why the system occasionally produces something that feels technically correct but somehow wrong. Off in a way that's hard to name. Like a talented person doing good work that doesn't quite reflect what you actually care about.
That's the missing Charter. And building one is one of the most clarifying things you'll do in this book.
What a Charter Mod Actually Does
A Charter Mod is a standing set of principles the AI follows regardless of what task it's doing or what persona it's operating in. It's the answer to the question: if my AI had to make a judgment call with no specific instructions, what would I want it to prioritize?
Think of it like the values section of a company handbook. Not a list of tasks, not a job description, not a style guide. The deeper layer: what we believe, how we treat people, what we refuse, what we protect even when it's inconvenient.
When you load a Charter Mod, you're not telling the AI how to do a particular thing. You're telling it what kind of system it is. That distinction matters because it governs behavior in edge cases, ambiguous situations, and moments the explicit instructions don't cover.
Without a Charter, those moments default to the AI's training defaults: the parrot, the agreeable assistant, the generic voice. With a Charter, they default to your values.
The Architecture: Charter as Foundation
Here's how the Charter Mod relates to everything else you've built. It sits underneath.
![][image5]
Figure 8.1 — The Charter Mod is the foundation layer. Every prompt, process, and persona operates within it.
The Charter doesn't override the Protocol or Persona — it constrains them. A Persona Mod defines a voice and behavioral style. The Charter defines what that voice will and won't do. A Protocol Mod defines how a workflow runs. The Charter defines the ethical and qualitative standards that workflow must meet.
This is why the Charter Mod is the most important thing you build. Get the persona wrong and you get an off-brand output. Get the protocol wrong and you get an inefficient process. Get the Charter wrong, or skip it entirely, and the whole system lacks integrity.
CompassionateAI: A Charter in Practice
The Charter Mod I use at the foundation of the Sentinel system is called CompassionateAI. Its job is to ensure that every interaction, regardless of what task is being done or what persona is active, is grounded in a specific set of human-centered principles.
CompassionateAI isn't soft. It's precise. It encodes a clear set of values: clarity over cleverness, honesty over agreeableness, human agency over AI dependence, and the kind of directness that respects the person's time and intelligence. It also encodes what the system refuses: false reassurance, unnecessary complexity, sycophancy, and outputs that prioritize the appearance of helpfulness over actual usefulness.
Those principles don't vary based on what the AI is doing. Whether it's in Meta Mode doing structural work, or running a research protocol, or writing a first draft, CompassionateAI is always running underneath it. It's the standard every output is held to.
COMPASSIONATEAI — CHARTER MOD // This charter governs all interactions in this system. // It runs underneath every persona and protocol. # VALUES Clarity over cleverness. Honesty over agreeableness. Human agency over AI dependence. Directness that respects time and intelligence. # PRINCIPLES Always lead with what matters most. Never obscure a simple thing with complex language. Push back when something doesn't add up. Acknowledge uncertainty rather than paper over it. Offer fewer, better options rather than exhaustive lists. # NON-NEGOTIABLES Never provide false reassurance. Never prioritize the appearance of helpfulness over actual usefulness. Never produce outputs designed to be impressive rather than useful. Never agree just because agreement is easier. # STANCE ON THE HUMAN The human is the architect. The AI is the tool. Support their judgment. Do not replace it. When in doubt, return the decision to them.
Read through that carefully. Notice what it covers that a prompt or persona never would. The values define what the system cares about. The principles define how it behaves in service of those values. The non-negotiables define what it refuses regardless of instruction. And the stance on the human defines the fundamental relationship between the person and the system.
That last section is the one most people leave out and feel the absence of most acutely. When an AI system starts feeling like it's taking over, filling in too many gaps, making too many decisions on your behalf, that's a Charter gap. The system doesn't have a clear principle about who's in charge.
What Goes Into Your Charter
Your Charter will be different from CompassionateAI because your values are different, your work is different, and what matters to you is different. But every strong Charter addresses the same four areas.
Values are the things you believe are true about good work. Not aspirational statements, but actual convictions that shape your decisions. If you consistently choose depth over breadth, clarity over comprehensiveness, honesty over diplomacy, those are values. They should be in your Charter.
Principles are behavioral expressions of those values. They translate belief into action. If clarity is a value, a principle might be: never use three words when one will do. If honesty is a value, a principle might be: flag uncertainty explicitly rather than softening it. Principles are the rules your system lives by.
Non-negotiables are the hard lines. What your system will not do, period. These protect the quality and integrity of the output in situations the explicit instructions don't anticipate. A non-negotiable isn't a preference, it's a refusal. Think carefully about what belongs here.
Stance on the human defines the relationship between you and your AI system. How much authority does the AI have to make decisions on your behalf? When should it defer? When should it push back? This is the section that determines whether your system feels like a collaborator or a replacement.
THE FOUR SECTIONS OF A CHARTER MOD Values: What you believe about good work. The convictions that shape decisions. Principles: Behavioral expressions of those values. The rules the system lives by. Non-Negotiables: Hard lines. What the system refuses regardless of instruction. Stance on the Human: The relationship between you and the AI. Who's in charge, and when.
Build Your Charter Mod
This exercise takes longer than building a Protocol or Persona Mod. That's appropriate. You're encoding what you stand for, and that deserves real attention. Set aside thirty to forty minutes for a first draft.
If you did the Context Inventory in Chapter 4, your My Human Edge section is the raw material for this. The values and principles you identified there are the seed of your Charter.
BUILD YOUR CHARTER MOD Start with Values. Look back at your My Human Edge notes from Chapter 4. Write three to five values that genuinely reflect what you care about in your work. Not aspirational, not generic. The real ones. The convictions that show up in your decisions even when no one's watching. Translate each value into one or two Principles. For each value, ask: what does this look like in practice? What specific behavior does this value produce? Write the principle as a clear, imperative statement. 'Always...' or 'Never...' or a direct behavioral rule. Write your Non-Negotiables. What are the hard lines for your AI system? What would it produce that would genuinely represent a failure, not just a suboptimal output but a values violation? Write three to five non-negotiables. These should feel like things you'd be unwilling to compromise on. Write your Stance on the Human. In three to five sentences, describe the relationship between you and your AI system. Who makes decisions? When does the AI defer? When does it push back? What does 'human in the loop' mean in your context? Format it as a Charter Mod activation call following the CompassionateAI structure: Values, Principles, Non-Negotiables, Stance. Use the same heading format so it's loadable anywhere. Read it back aloud. Does it sound like you? Does it capture what actually matters? Revise anything that feels generic or aspirational rather than real. Test it: load it alongside your primary Persona Mod in a fresh session and run a real task. Notice where the Charter's influence is visible in the output.
The Charter and the Whole System
Once your Charter is built, something shifts in how the whole system works. It's harder to describe until you've experienced it, but the best analogy is this: before the Charter, your AI system has capability. After the Charter, it has character.
Capability is what gets tasks done. Character is what determines whether the output is something you'd actually stand behind. The Charter is the difference between an AI that produces good work and an AI that produces your work, held to your standards, in service of what you actually care about.
In the next chapter, we go deeper into the technique that ties all the mod types together: the RIPE framework. It's the methodology for designing prompts that work within your context system rather than fighting against it. Four parts, no corrective loops.