XiaoyaoClaw Home
Give every OpenClaw agent a "home": standard directory layout + WORKSPACE.md spec + safe multi-agent config — no messy folders, no lost memory, no config collisions.
Product Introduction
OpenClaw Workspace Initializer is an OpenClaw skill that solves three agent-workspace pains in one shot: directory layout, persistent spec, and multi-agent config safety. Every OpenClaw session starts fresh — without a spec, agents dump files at the root, forget their memory/, or clobber openclaw.json via config.apply. This skill auto-creates the standard projects/ tasks/ outputs/ knowledge/ scripts/ memory/ tmp/ directories on first entry (or reset), writes WORKSPACE.md (so the spec survives restarts), detects multi-agent setups and persists the "config.patch ✅ / config.apply ❌" rule into AGENTS.md, and logs the init into memory/. Battle-tested on 7 agents sharing one openclaw.json.
Core Features
- Standard layout — auto-create projects/, tasks/, outputs/, knowledge/, scripts/, memory/, tmp/
- Persistent spec — WORKSPACE.md survives restarts; agents read it before acting
- Multi-agent safety — "config.patch ✅ / config.apply ❌" rule persisted into AGENTS.md to avoid clobbering shared config
- Memory preserved — init log written to memory/ so agents never lose context after restart
- Reusable templates — templates/ (WORKSPACE.md + multi-agent safety spec) can be reused standalone
- Battle-tested — born from a real production incident where config.apply wiped 7-agent config
- Easy install — one ClawHub command (clawhub install xiaoyaoclaw-workspace-initializer) or manual from GitHub
Use Cases
Screenshots
Step 1 — connect your agent
Create and connect a new agent (Feishu Bot / WeChat / Telegram, etc.) — set its name, title, and responsibilities