Hq Starter — Development skill for Claude Code

HQ

HQ is a file-based personal memory system for a coding agent, Claude Code, Codex, or anything similar. Point the agent at this repo, answer some questions about yourself, and you get an agent that remembers who you are, what you're working on, and how you like to work, across sessions, without a database, a hosted service, or an app to maintain.

Under the hood it's markdown and JSON files in a git repo, plus one small script the agent calls to write them.

Contents

Three ways in, depending on who you are:

• "Just set me up." Go straight to Quick start. It's one pasted prompt, and it's written so that someone who has never opened a terminal can follow it.
• "Convince me first." Read What this gets you, and if you're weighing HQ against something specific (Obsidian, ChatGPT's memory, OpenClaw, Hermes), see guides/comparisons.md. The reasoning behind the design is in How it works.
• "I already set it up." Head to After setup for the first-week guide and the upgrades: apps, proactivity, phone access.

All sections:

1. What this gets you: why bother, in terms of things you can actually ask for
2. Quick start: three steps, one pasted prompt
   • If you've never used a terminal: reassurance and a five-term glossary
3. What setup creates: the folder you end up with, file by file
4. After setup: the guides, from first week to apps, proactivity, and phone access
5. How it works: the three-layer design and why it's files instead of a database
6. Honest caveats: what this is not

What this gets you

A regular Claude or ChatGPT conversation starts over every time, and the built-in "memory" features hold a short list of facts that the vendor's system decides to keep. An HQ changes what you can actually ask for: