Skip to content

Prompt Playbook

This page explains how prompts are organized in the YoMemo docs. As the number of prompts grows, we use a three-layer structure and a uniform prompt shape so they stay easy to find and reuse.

LayerMeaningExample
Protocol / capabilityAbstract rules; one protocol can power many scenarios. Few and stable.Advanced Memory Engine & Ontology Protocol (v0.0.0)
Scenario templatesOne doc per scenario with copy-paste blocks (input → behavior → output).Cursor proactive memory, Advanced Memory Engine scenarios
Recipes (Cookbook)One-off, concrete examples (e.g. “this week’s Cursor prompt”); can live in blog or examples.Ad-hoc prompts you use for a specific setup

Protocols are the single source of truth: scenario docs reference or trim them instead of duplicating long blocks, so versions don’t drift.

When adding any new prompt, split it into three blocks for easier maintenance and composition:

  1. Role — Who the model is (e.g. “High-Order Memory Processing Engine”).
  2. Behavior / Rules — What it does (checklist for protocols; “input → process → output” for scenarios).
  3. I/O contract — What you give it (context, memories, user message) and what it returns (structured fields, append ✓, etc.).
  1. Choose the layer — Is it a protocol, a scenario template, or a one-off recipe?
  2. Name and handle — Pick a filename and (if you manage it in YoMemo) a handle, e.g. prompt-scenario-weekly-review.
  3. Write the three blocks — Role, Rules, I/O contract.
  4. Version and tags — Set vX.Y and which LLM/tool it’s for (Cursor, Claude, GPT, etc.).
  5. Document or not — Protocols and main scenarios go into this site; experimental or personal prompts can stay in YoMemo only.

Raw prompt text lives in the dedicated prompts repo
https://github.com/yomemoai/yomemo-prompts
For the Advanced Memory Engine protocol, see:
https://github.com/yomemoai/yomemo-prompts/blob/main/system-v0.0.0.txt. Scenario pages in the docs reference and trim those protocols.