OpenClaw Multi-Instance

OpenClaw 多实例互联。当用户有多台机器运行 OpenClaw，想让它们互相通信、传文件、共享记忆时使用。触发词：多实例、互联、另一台服务器、同步记忆、实例通信。

Security Scan

The skill's files and runtime instructions are internally consistent with its stated purpose (linking multiple OpenClaw instances for API communication, scp transfers, and periodic memory merging); it requires user-provided tokens/SSH keys and performs obvious file sync operations, so the main risk is granting full control to paired instances rather than covert or unexpected behavior.

Purpose & Capability

The skill name/description (multi-instance communication, file transfer, memory sync) matches the included SKILL.md, README, and sync script. The provided script and templates are directly relevant to enabling API calls, SSH/scp transfers, and merging MEMORY.md and per-day memory files.

Instruction Scope

The SKILL.md instructs the agent/user to enable a remote OpenClaw API endpoint, obtain gateway tokens (gateway.auth.token or OPENCLAW_GATEWAY_TOKEN), set up SSH key-based auth, use scp to copy identity/memory files, and create cron jobs for periodic sync. Those actions are within scope for multi-instance linking, but the instructions reference sensitive artifacts (API tokens, SSH keys, TOOLS.md may include API keys) and propose operations that grant significant control (e.g., adding authorized_keys via API). The skill does not attempt to read unrelated host files or call unexpected external endpoints.

Install Mechanism

This is an instruction-only skill with no install spec. No third-party downloads or extract operations are present. The only executable file is a local shell script included in the repository, so nothing is pulled from external URLs during install.

Credentials

The manifest declares no required env vars, but the documentation explicitly mentions gateway.auth.token / OPENCLAW_GATEWAY_TOKEN as the API credential to use. That is proportional to the stated purpose (remote API authentication), but users should be aware those tokens provide full control of the remote Gateway. No other unrelated credentials are requested by the skill.

Persistence & Privilege

The skill does not request always:true or other elevated platform privileges. It suggests creating a cron job on the user's systems (a normal, explicit local change) but does not modify other skills or system-wide agent settings. Autonomous model invocation is allowed by default but not excessive here.

Assessment

This skill appears to do what it says: connect OpenClaw instances via the remote API and SSH and merge memory files. Before installing or running it, be mindful: (1) API tokens (gateway.auth.token / OPENCLAW_GATEWAY_TOKEN) grant full remote control — treat them like root credentials; do not share them with untrusted hosts. (2) Setting up passwordless SSH or instructing a remote instance to append your public key grants file-transfer and remote-execution ability — prefer a non-root user, use passphrase-protected keys, and consider an overlay network (Tailscale) instead of exposing ports. (3) Inspect IDENTITY.md / TOOLS.md and memory files before syncing because they may contain API keys or secrets that would be copied. (4) Test on throwaway instances first and back up local MEMORY.md and memory/ before running the sync. The skill's merge logic is heuristic and may duplicate or append content; review results after the first run. If you want extra assurance, require that each sync step be manual (no cron) and limit accounts used for ssh/api to minimal privileges.

OpenClaw 多实例互联

让多台机器上的 OpenClaw 实例互相通信、传文件、共享记忆，实现无感切换。

什么时候用这个 Skill

• 用户说"我还有一台服务器也跑了 OpenClaw"
• 用户想从一台 OpenClaw 操控另一台
• 用户想让多台实例共享记忆/身份
• 用户提到多实例互联、同步等需求

架构概览

┌──────────────────┐         ┌──────────────────┐
│   实例 A (本地)    │         │   实例 B (远程)    │
│                  │  SSH    │                  │
│  OpenClaw    ────┼────────▶│  OpenClaw         │
│                  │  API    │                  │
└──────────────────┘         └──────────────────┘
        │                            │
        └──── 定期对账（记忆叠加合并） ──┘

三种能力：

1. API 通信 — 实例之间直接对话
2. SSH 传文件 — 免密传输任意大小文件
3. 记忆同步 — 定期合并记忆，不覆盖，只叠加

搭建流程

Step 1: 开启远程实例的 API Endpoint

远程实例需要在配置中开启 Chat Completions endpoint：

// ~/.openclaw/openclaw.json
{
  gateway: {
    http: {
      endpoints: {
        chatCompletions: { enabled: true },
      },
    },
  },
}

然后重启 Gateway：openclaw gateway restart

验证：用 curl 测试：

curl -sS http://<远程IP>:<端口>/v1/chat/completions \
  -H 'Authorization: Bearer <TOKEN>' \
  -H 'Content-Type: application/json' \
  -d '{"model":"openclaw:main","messages":[{"role":"user","content":"ping"}]}'

获取 Token：查看远程实例的 gateway.auth.token 或环境变量 OPENCLAW_GATEWAY_TOKEN。

Step 2: 配置 SSH 免密登录

如果本地有公钥，直接让远程实例帮忙加：

通过 API 告诉远程实例：
"请把以下公钥加到 authorized_keys：
<本地 cat ~/.ssh/id_ed25519.pub 的内容>
追加到 ~/.ssh/authorized_keys，chmod 600"

如果远程有公钥，本地加：

# 获取远程公钥（通过 API 问远程实例要）
# 然后本地：
echo "<公钥内容>" >> ~/.ssh/authorized_keys

测试：ssh <用户>@<远程IP> "echo OK"

Step 3: 同步核心文件

通过 scp 把身份和记忆同步过去：

scp IDENTITY.md SOUL.md USER.md MEMORY.md TOOLS.md AGENTS.md \
    <用户>@<远程IP>:<远程workspace路径>/

scp -r memory/ <用户>@<远程IP>:<远程workspace路径>/memory/

注意权限：远程文件要 chown 给正确的用户。

Step 4: 设置定期记忆同步

创建 cron job，定期执行记忆合并。

合并原则（核心）：

• 叠加，不覆盖 — 只增不减
• MEMORY.md：对比两边，互相补齐对方没有的内容段
• memory/ 日记（YYYY-MM-DD.md）：按日期补齐缺失文件，同一天的去重合并
• 冲突内容保留两份，标记来源 <!-- 来源: 实例A -->

Cron 配置建议：每周一次，比如周日 00:00

sessionTarget: isolated
schedule: { kind: "cron", expr: "0 0 * * 0", tz: "Asia/Shanghai" }
payload: {
  kind: "agentTurn",
  message: "执行记忆同步：<1> scp 拉取远程的 MEMORY.md 和 memory/ <2> 叠加合并 <3> 推送合并结果回远程 <4> 输出同步报告"
}

使用方式

搭建完成后，跟本地实例说：

• "让服务器做 XXX" → 本地通过 API 转达给远程
• "把这个文件传到服务器" → scp 直传
• "查一下服务器的 XXX" → API 转达

跟远程实例说也一样，体验无感切换。

文件说明

本 skill 包含：

文件	作用
SKILL.md	本文件，给 OpenClaw 的搭建指南
scripts/sync-memory.sh	记忆同步脚本（可选，也可以让 agent 自行合并）
references/peer-config.json5	远程实例连接信息模板

安全提醒

⚠️ 必须告诉用户：

• API Token = 完全控制权，不要泄露
• 公网 IP 的实例建议用 Tailscale，不要直接暴露
• SSH key 建议加密码保护
• TOOLS.md 里可能有 API key，同步时注意安全