Prompt contracts
The Universal Prompt Contract: intent, context, route, output shape, constraints, verification, and reuse policy. Ambiguity removed before generation starts.
Prompt Operations Manual · v2.3 release candidate
The operating manual for
AI-assisted work.
Treat prompts, agent handoffs, templates, tool use, and verification as engineering assets — named, versioned, reviewed, and released. Built for builders, engineers, SEO operators, and technical teams.
intent
context
contract
route
output
verification
release
Overview
From one-off prompts to a repeatable operation
The unit of work is not a clever prompt. It is a prompt that produced a useful output under known constraints, with enough evidence that someone else can reuse it. The Manual is the operating layer that makes that repeatable — so prompt work stops disappearing into chat history and starts behaving like versioned engineering.
The system
An operating system, not a list of tips
Template library
Reusable, versioned templates with registry metadata and a quality bar — a prompt shelf, not a pile of lost transcripts.
Before / after rewrites
Worked examples that turn a vague request into an operational one — and explain why the rewrite holds up under review.
Review gates
Human checkpoints that stay real instead of decorative. If you can’t explain why an output is safe to ship, it isn’t ready.
Agent, skill & tool contracts
The same discipline extended to agents, skills, and tools — so capabilities compose predictably instead of drifting.
Role preambles & handoffs
A 12-role catalog and lifecycle handoff map so work moves cleanly from strategist to researcher to builder to reviewer to release.
Audience
Who it is for
Solo operators
You want AI output you can trust and reuse — not regenerate from scratch every time.
Engineers & technical leads
You’re building AI-assisted workflows and need contracts, routes, and verification you can hand to a team.
SEO & web practitioners
You treat prompts as versioned engineering artifacts with audit trails, not disposable one-offs.
Small teams
You need lightweight review discipline instead of trusting whatever sounds most confident.
Contents
What is inside
- The prompt operations model and the operating loop
- Universal Prompt Contract and output typing
- Domain template library with registry metadata
- Before / after prompt rewrites with rationale
- Workflow schematics and execution routes
- Training and human-in-the-loop review
- Metrics, governance, and risk tiers
- Adoption roadmap and prompt maturity model
- Agent, skill, and tool operations
- Role preambles and lifecycle handoffs
- Appendices: review worksheet, eval-case record, Role Lens prompts
- A look ahead at the Cathedral Workstation Edition
Editions
Choose how you use it
Operator Edition
Coming soonThe full v2.3 Manual as a single, versioned PDF reference. The direct / operator edition.
Free Toolkit
Coming soonTemplates, worksheets, checklists, and schemas you can put to work immediately.
Cathedral Workstation
LaterA full local, repo-aware operating environment — memory, model routing, and review gates. A future edition.
Release discipline
A versioned, living manual
This is not a one-and-done ebook. Updates ship as numbered releases with a public changelog, so the artifact never quietly drifts. The current direction is the v2.3 operator edition.
Early access
Be first when it ships
Want the Operator Edition or the Free Toolkit the moment they land? Send a note and you’ll be on the list.
Contact address is a placeholder for launch. No payment or downloads are live yet.