stem
grandamenium
A Claude Code workspace that interviews you, then builds itself around your actual work.
Open grandamenium/stem → Checked against the four rules.
Built for Claude Code
ships .claude CLAUDE.md and nothing for other agents.
In its own words
Skills, memory, projects, and scheduled agents that maintain themselves.
A workspace that interviews you, then builds itself around your actual work.
Most "AI setup" repos hand you a folder of prompts and wish you luck. Stem asks you what you do all day, connects your tools, turns your repetitive work into skills, and puts the ones that can run on a schedule on autopilot. Then it keeps itself organized as your work grows.
Quickstart
Read
.claude/skills/onboard/SKILL.mdand onboard me.
Set aside 70 to 90 minutes, or about two hours if you want it done thoroughly. You can stop anywhere and pick up later.
What you get
By the end of onboarding your workspace contains:
What makes it different
It maintains itself.
Every night a scheduled run reads what happened that day, promotes what matters into long term memory, deletes what turned out to be wrong, and fixes skills that misbehaved. Every week another run reconciles every index against what is actually on disk, finds stale tasks and broken links, and tells you if any of your automation has drifted.
That is the difference between a system that gets better as you use it and a folder of prompts that quietly rots.
What runs by itself, and what does not
This is the honest part, and it is worth understanding before you start.
Claude Code can run scheduled agents, and those run in the cloud. That is genuinely good: they work while your laptop is closed. But it also means they cannot touch your machine.
The repository's own description, then 57% of its README. Read the rest on GitHub ↗
Read from the repository's git tree on 2026-08-17.