Claude Space Self-Ideator: Using Claude to Brainstorm Claude Workspaces

The Project

https://github.com/danielrosehill/Claude-Space-Self-ideator

One of the most interesting things about the Claude Space pattern is how naturally it lends itself to recursion. If a version-controlled repository can serve as a focused workspace for human-AI collaboration on any domain, then why not use one to brainstorm ideas for more of them? That is exactly what the Claude Space Self-Ideator does: it is a Claude Space whose entire purpose is generating concepts for new Claude Spaces.

What Is a Claude Space?

For those unfamiliar with the pattern, a Claude Space is a version-controlled repository that functions as a miniaturized agentic workspace. It uses folder structure, system prompts (via CLAUDE.md), slash commands, and sub-agents to create a focused environment for human-AI collaboration on specific domains or problems.

The core insight is that a code repository's folder structure can serve as a simple, transparent, and portable way to manage AI context. No databases, no embeddings, no complex infrastructure required. Git provides the version control. Markdown files provide the context. And the repository boundary provides the scope.

How the Self-Ideator Works

The Self-Ideator repository contains a structured collection of Claude Space concepts organized by category, along with examples of existing working spaces and documentation on the pattern itself. Running the /develop-idea slash command with a concept generates a fully structured idea document following a standard template, complete with suggested folder structure, workflows, slash commands, and agents.

The repository already contains ideas spanning a wide range of non-technical domains:

• Personal Development — Habit Laboratory, Life Audit, Values Compass

• Learning & Skills — Learning Lab, Skill Tracker

• Creativity & Projects — Idea Incubator, Creative Project Manager

• Home & Life Management — Home Operations, Meal Intelligence

• Mental & Emotional Wellbeing — Anxiety Workbench, Gratitude Practice

• Specialized Domains — Travel Planner, Garden Intelligence, Pet Care, Emergency Preparedness

Why This Pattern Matters

The Self-Ideator is a demonstration of something I find genuinely powerful about Claude Spaces: they are domain-agnostic. The same pattern that works for systems administration or code generation works equally well for tracking habits, planning meals, or managing a garden. The repository provides approximate boundaries for agents and humans to work within, achieving what otherwise requires complicated infrastructure.

Using Markdown files as context provides a lightweight alternative to formal vectorization or retrieval-augmented generation. And GitHub's template structure allows significant upfront investment in workflow design that can be reused whenever a suitable project arises. Each idea document includes a difficulty rating (easy, medium, or hard), a duration type (ongoing or time-limited), and a complete specification for folder structure, slash commands, and agents.

The repository also includes documentation on agent portability, noting that these patterns are not locked to Claude Code and can work with other agentic CLIs like Codex or Aider. The six-step development pattern documented in the repo provides a repeatable formula for building new Claude Spaces from scratch.

If you are curious about non-code applications of agentic AI, this is a good place to start exploring.