Erin Stanley, Evoked·Restraint spec workshop

Your agent will do something you never told it to.

Let's decide what it must refuse.Before it does.

A focused, fixed-fee workshop that produces an opinionated restraint specification for your actual agents: what they may do, what they must refuse, and how you verify it. Run by the founder of a live 143-agent governance system.

Buy the workshop - $2,500 Book the 20-minute call first →

Who this is for

You lead a team shipping an AI agent or assistant, and one of these is true.

You have a shipping deadline and nobody owns "what should this thing refuse."
A customer, a regulator, or your board has started asking how the agent handles data and edge cases.
You had a near-miss - an output, a data path, a tool call - that made you realize the guardrails are implicit, not written down.

If your agent is a weekend prototype, you do not need this yet. If real users touch it, you do.

What we do together

A single working session, half-day and remote, plus a written deliverable. You bring your agent and how it actually behaves. We produce, together, three things you keep.

A restraint spec for your system - the specific things your agent must not do, written as a document your engineers can build against and your reviewers can check.
A refusal map - where each boundary lives, what triggers it, and what happens when it fires.
A verification checklist - how you confirm the agent actually holds the line, not just that you intended it to.

You leave with a document, not a slide deck. It is yours to keep, extend, and reuse.

What this is not

Not software. There is nothing to install and no new dependency.
We never touch your customer data. This is advisory. Your data stays in your systems. That is a deliberate design choice, and it is why we can move in days instead of quarters.
Not a generic template. The spec is written for your agent, your risks, your users.

Why me

Operator

A live 143-agent governance system in daily operation - identity, memory, refusal rights, dissent records. Restraint architecture I run, not a theory I read about.

Patent

Inventor on active filings in AI agent governance. I publish restraint and sovereignty specifications openly.

Principle

We evoke, we never extract. No dark patterns, no dependency, no harvesting. If a boundary protects your users, I help you draw it and hold it.

I am not a large firm with a compliance stamp. I am the person who has actually specified this, for a system larger than yours, and can do it for yours in a focused session.

Price

Restraint spec workshop $2,500

One workshop. One written spec. One verification checklist.

Half-day, remote. Fixed fee, no hourly meter, no upsell required.

If, after our first 20 minutes, this is not the right fit, we stop and there is no charge.

Buy the workshop - $2,500 Book the 20-minute call first →

Next step

Book a 20-minute call first. I want to hear what you are shipping and whether a restraint spec would actually help you this quarter. If it fits, we schedule the workshop. If it does not, I will tell you.

Book the 20-minute call Or buy the workshop now - $2,500

Your agent already has boundaries. Right now they live in prompts and code, implicit, unverified, owned by no one. A restraint spec is where they become a document your team can build against and your reviewers can check. If real users touch your agent, that document is worth an afternoon.

Erin Stanley, Evoked