← ClaudeAtlas

nomadlisted

Turn any folder into a portable, multi-harness AI team workspace — build a new team from scratch, migrate an existing scattered setup into a clean skeleton, or mount an already-built team onto a new AI harness (Claude Code, Codex/ChatGPT, OpenCode, Copilot, Cursor). Trigger on requests like "make my AI team portable", "give my agents persistent memory across tools", "set up a multi-agent team folder", "migrate my prompts/config into a team structure", "connect this team to another AI tool", "sync my agent across tools", "AI team memory portability", or "switch AI tools without losing my agent's memory".
LucioLiu/nomad · ★ 1 · AI & Automation · score 64
Install: claude install-skill LucioLiu/nomad
# Nomad — Portable AI Team Workspace Nomad teaches an AI agent to turn a plain folder into a **portable, multi-harness team workspace**: a folder that, when opened by any capable AI tool, self-configures that tool with the right personality, memory, and collaboration rules — no cloud service, no proprietary format, just files. This is not a product to install. It is a **skill**: read it, then act on the user's folder directly. **Terminology note**: the generic term for "an AI tool/CLI/IDE capable of reading files and running an agent session" is **harness** throughout this skill and its reference files — Claude Code, Codex/ChatGPT, OpenCode, Copilot, and Cursor are all harnesses. **Platform** is reserved for passages describing one *specific* harness's own behavior — e.g. the per-platform table in `mounting-guide.md` §3. If the two appear used loosely elsewhere in these files, treat them as referring to the same concept. ## Step 0 — Ask the user's language, always Before doing anything else, ask (or infer from how the user is writing to you): > "Would you like to proceed in English or 中文?" Every reference file in this skill has two parallel versions — `references/<name>.md` (English) and `references/<name>.zh-CN.md` (Chinese). Once the language is known, **read only the matching version** for the rest of this task. The two versions are content-equivalent; never mix languages mid-task, and never presume English by default just because this file is in English. ## Step 1