Operator Reference · v.2026.4.2

Stop re-explaining your project to Claude every session.

The operator's manual Anthropic doesn't ship — a protocol, an interactive network, and three drop-in templates for solo founders and small teams who feel session-to-session friction with AI assistants.

16 sections 5 worked examples 3 templates One-page cheat sheet
§ Diagnose

Five symptoms. If three match, this is for you.

Most "how to prompt Claude" guides solve the wrong problem. The pain isn't phrasing. The pain is operating without discipline across sessions, projects, and surfaces. Read these symptoms before the pitch.

01
You re-explain your project context at the start of most sessions.
02
You've lost track of a decision you and Claude already settled, and re-litigated it later.
03
You've gotten output you'd describe as "fine but generic" when you wanted something specific to your work.
04
You finish a long session sensing the last hour produced less than the first.
05
You have two or more half-built projects with Claude where you've stopped being able to remember what state they're in.
Three or more matches? Continue. The protocol gives you something usable. Two or fewer? Bookmark this and come back in a month. The discipline only makes sense once you've felt the failure mode it prevents.
§ Bundle

Two artifacts, three templates, one cheat sheet.

A system, not a stack. Each piece has a job — the protocol breaks if you remove one.

Primary
The Protocol
A 16-section interactive reference. Qualifier opener (filter yourself in or out). Day 1 / 7 / 30 reader track. Worked examples on the highest-traffic sections. Three operator profiles in the metrics layer (illustrative numbers, real failure modes). 12-item self-audit at the end with cadenced triggers. Audit state persists across sessions. Export to dated markdown. Print-mode cheat sheet built in.
Single HTML · Light/dark · Works offline · ~25 min first read
Interactive
The Network
The same content as a layered network. 16 nodes, 3 tiers, real dependencies as edges. Map mode + Book mode. Per-node audit (48 items, finer-grain than the protocol's 12).
Single HTML · Cross-tab sync
Drop-in
Three Templates
CLAUDE.md, CLAUDE_JOURNAL.md, SKILL.md. Each annotated with what goes in / what doesn't. Saves 30 minutes per new project.
Markdown · Drop into any repo
Physical
One-Page Cheat Sheet
Print double-sided, keep at desk. Three columns: surface routing, prompt patterns, top metrics. Until the protocol is muscle memory, the cheat sheet is the reminder.
PDF · US Letter · Printable
Maintained
Public Changelog
Semver versioning. Updates when a failure mode emerges in operator practice. Re-download to get the current version.
Markdown · Versioned
§ Adoption

Day 1, Day 7, Day 30.

The protocol is a system. Reading it linearly takes about 25 minutes. Living it takes weeks. This track orders what to install when, so you build discipline before you exhaust enthusiasm.

Day 1 · Setup
Install the foundations
  • Read the Premise and Surface Routing sections
  • Drop CLAUDE.md and CLAUDE_JOURNAL.md into your project
  • Write your first journal entry — even a one-liner
  • Bookmark the protocol. Don't try to memorize.
Day 7 · Practice
Add the discipline layer
  • Read Prompt Patterns and Build Sessions
  • Run one decision through the multi-pass pattern
  • Write an out-of-scope list before a build session
  • Run the self-audit. Expect a low score.
Day 30 · Measure
Add the measurement layer
  • Read Metrics and Cadence
  • Run your first weekly review — keystone
  • Write a skill for the most repetitive domain task
  • Re-run the audit. Compare to Day 7.
Team license
For up to 10 operators
$149
One-time · Same bundle · One contact for support
Everything in Solo, plus
  • Up to 10 operators in one organization
  • One contact for billing and support
  • Free updates within major version
  • Internal sharing license included
  • Email support · 48-hour SLA
  • Save $341 vs 10 individual licenses
Get team license
7-day refund · No questions asked

Questions worth asking.

Is this another prompt engineering guide?

No. Prompt phrasing is a small lever. The protocol is about operating discipline across sessions, projects, and surfaces. If you've already read Anthropic's prompt engineering docs, this is the layer they don't cover — what to do between prompts.

Does this work for Claude only, or also ChatGPT / others?

The protocol is generic. The surfaces (Web chat, Claude Code, Cowork, API) are Claude-specific names, but the routing principle applies to any assistant ecosystem with multiple surfaces. The journaling, audit, and cadence layers are platform-agnostic.

If you primarily use a different assistant, expect to rename the surfaces. Everything else applies.

Why $49 and not free?

The free preview gives you the philosophy — qualifier, reader track, and most of the 16 sections. That's enough to decide whether the discipline maps to your work.

The paid bundle gives you the implementation — the working audit ritual, worked examples, operator profiles, templates that drop into a real repo, the printable cheat sheet, and the interactive network. Paying $49 is a commitment device. People who pay run the audit. People who download free guides read once and close the tab.

How is this different from a Notion template or Obsidian vault?

Notion and Obsidian are tools for storing notes. The bundle is a set of artifacts for governing how you work with an AI assistant. Different layer.

That said: if you operate primarily out of Notion or Obsidian, the three templates and the protocol content port cleanly. You'd be giving up the interactive audit and the network visualization, which is the trade-off.

Will it become outdated?

Anthropic ships features fast. Sections referencing specific surfaces (Cowork, Claude Code, API) may need updates as the products evolve. The protocol commits to a semver-versioned public changelog — re-download to get the current version. Major updates within the same year are included.

The discipline layer — journaling, audit, cadence — does not change. Those principles long-outlive any specific tooling.

Is this for individuals or teams?

Designed for individuals first. The protocol scales to small teams (2–5 people) by treating each team member as an operator and using shared CLAUDE.md plus a shared journal. For larger teams, the discipline still applies but you'll want to add team-coordination layers the protocol doesn't cover.

Can I share the bundle with my team?

Personal use is one purchase per operator. If you want to roll this out to a team of more than a few, reach out about a team license — we can work something out that doesn't feel punitive.

What's the refund policy?

7 days, no questions. If the protocol doesn't fit your work, the friction wasn't real to you yet — that's fine, refund and revisit later.

The friction is real. Or it isn't.

If three or more symptoms matched, the protocol gives you something you can install this weekend. If they didn't, save the URL for when they do.