# House Rules — [Your Business]

> One page. Rules enter ONLY through the weekly review — never directly. Every rule carries its date and the incident that taught it. This page, not the ledger, is what your AI tools read.

## The rules

*(Starts empty. That's correct. Rules arrive through the ritual.)*

- **[YYYY-MM-DD]** [The rule, stated so it can be violated — "never quote a market stat without a named source," not "be careful with numbers."] *(source: [ledger row / incident])*

**Retiring a rule:** strike it through with a date and a one-line reason. Never silently delete — the graveyard is where new hires learn why the current rules exist.

---

## Loading this page into your tools (do this once)

The closing move: every AI conversation starts with these rules already loaded — automatically, not "remembered."

**The instruction to attach wherever your tool takes persistent instructions:**

> Before any task, read the house-rules page and follow every rule. If work in this session teaches a lesson that should change future work, say so at the end and suggest a ledger row.

*(That second sentence gets you prompted capture — the AI drafts claims, a human still files them.)*

### ChatGPT — solo
Settings → **Customize ChatGPT** → paste your rules into the free-text field. Budget: ~1,500 chars free tier, ~5,000 paid. Hitting the wall is a healthy sign — graduate to a Project and run a retirement pass.

### ChatGPT — team
One shared **Project**. Upload this page + the ledger to it; set the **Project instructions** to the instruction above. Project instructions apply to every conversation inside and override personal settings.

### Claude — solo
Two slots, use both: **Cowork** → Settings → Global Instructions (paste rules); **claude.ai** → a Project with the rules in project instructions and this doc + ledger as project knowledge.

### Claude — team
One shared **Project**: rules + ledger as knowledge, the instruction above as project instructions, visibility **Public** (org-wide), staff get **"Can view"**, the weekly-ritual owner gets **"Can edit."** That permission split IS the approval gate.

### Mixed shop / any other tool
Keep exactly one rules page and one ledger in a place both camps open (Google Drive works — both ChatGPT and Claude have Drive connectors that read the live doc). Each tool loads the same page through its own mechanism. Any tool with no instructions slot: pin this doc where the tool (or its operator) reads it at session start.

**Why per-tool "memory" features are not this page:** memory is the model deciding what to remember about you — unshared, unapproved, per-account. Useful ambient context; structurally wrong for house rules.