claude-workspace-public

A self-managing Claude Code workspace built over weeks of daily use. It handles task management, research, content digestion, travel planning, and self-improvement, all orchestrated through skills, hooks, and a custom task engine.

It's designed to be cloned and used: point your Claude at this repo, and it can walk you through setup, explain how everything works, and help you customize it. That's the point: if a system is built with agents in mind, the agent can do the knowledge transfer.

Key Concepts

Skills

Structured behaviors that fire on demand. Each skill has a SKILL.md that defines its trigger, process, and conventions. Skills range from simple (task management) to complex (travel search with multi-source comparison). They're the primary way the workspace extends Claude's capabilities.

Skills in this workspace: task, brainstorm, research, digest, travel, shop, process, ingest, autonomous, skill-creator, using-skills.

Hooks

Automatic injection points. The SessionStart hook loads the skill enforcement system on every new conversation. The UserPromptSubmit hook injects per-prompt rules (task gates, brainstorm gates, improvement logging) before every response. Together, they ensure Claude follows the workspace's conventions without being asked.