OpenClaw Manager
Overview
Build and operate OpenClaw with production-safe defaults across both local and hosted environments. This skill is optimized for operators with limited platform expertise and enforces hard security gates before rollout completion.
Primary references:
- - Docs map: INLINECODE0
- Security gates checklist: INLINECODE1
- Mode matrix: INLINECODE2
- OS matrix: INLINECODE3
- Integrations playbook: INLINECODE4
- Ops ledger schema: INLINECODE5
Automation helpers:
- - INLINECODE6
- INLINECODE7
- INLINECODE8
Default ops ledger path:
- -
./openclaw-manager-operations-ledger.md (or operator specified)
Hard-Stop Rules (Never Bypass)
Stop and block deployment/install progression if any condition is true:
- 1. Required secrets profile fails validation.
- Security checklist mandatory gates are not all passing.
- Rollback path is not documented and owned.
- Ops ledger was not updated for the current phase.
- Public exposure requested without auth boundary and token controls.
Workflow
1) Intake and Scope Lock
Collect and confirm:
- -
mode: local or INLINECODE12 - INLINECODE13 :
local, fly, render, railway, hetzner, INLINECODE19 - INLINECODE20 :
macos, linux, INLINECODE23 - INLINECODE24 : subset of
telegram, discord, INLINECODE27 - INLINECODE28 : subset of
email, INLINECODE30 - INLINECODE31 :
dev, staging, INLINECODE34 - INLINECODE35 :
private or INLINECODE37
Before proceeding, write a scope_lock ledger entry:
CODEBLOCK0
2) Generate a Decision-Complete Plan
Always generate a plan first:
CODEBLOCK1
The plan output is the execution contract. Do not skip sections.
3) Validate Secrets and Config Profile Before Any Infra Change
Validate environment using profile-aware gates:
CODEBLOCK2
Validation enforces:
- - required keys by profile
- required provider/model alternatives
- malformed and duplicate env keys
- placeholder values
- weak gateway/setup tokens
- legacy alias warnings
Write a predeploy_validation ledger entry immediately after validation.
4) Execute Mode Branch
Branch A: Local install (mode=local)
- 1. Use official install/onboarding docs for local setup.
- Apply OS-specific commands from
references/openclaw-os-matrix.md. - Validate startup, persistence path, and local auth boundaries.
- If local public exposure is requested, apply gateway hardening gates from security checklist first.
Branch B: Hosted clone + deploy (mode=hosted)
- 1. Clone the selected OpenClaw source repo.
- Follow provider playbook from
references/openclaw-doc-map.md. - Configure persistent storage before production traffic.
- Configure ingress/auth, secrets, and health checks.
- Verify runtime logs for startup/auth errors and secret leakage.
Write a deploy_complete ledger entry once deployment/install is complete.
5) Configure Channels and Integrations Safely
For each selected channel/integration:
- - inject credentials via secret manager/env only
- run a minimal smoke test
- verify auth boundaries and error logging safety
Track each item as:
- - INLINECODE45
- INLINECODE46
- INLINECODE47
Use references/openclaw-integrations-playbook.md for email/calendar specifics.
6) Agent + Memory Baseline
Document and validate:
- - memory persistence strategy
- retention expectations
- restart and recovery behavior
- agent behavior boundaries
Update ledger with operational baseline decisions.
7) Mandatory Security Gate
Run
references/openclaw-security-checklist.md and produce pass/fail per gate.
No go-live if any mandatory gate fails.
Write a security_gate ledger entry with explicit blockers (if any).
8) Handover and Incident Readiness
Produce:
- - provider status summary
- channel + integration matrix
- security gate table
- rollback and escalation ownership
- follow-up actions by risk order
Write handover ledger entry. For incidents/troubleshooting, append incident entries as events happen.
Output Contract
Always return:
- 1. Mode/provider/OS/environment status summary
- Security gate results (hard pass/fail)
- Channel + integration matrix
- Agent + memory configuration summary
- Ops ledger update confirmation (event names written)
- Follow-up actions ordered by risk
OpenClaw Manager
概述
在本地和托管环境中以生产安全默认值构建和运行OpenClaw。该技能针对平台专业知识有限的运维人员进行了优化,并在完成部署前强制执行严格的安全门控。
主要参考文档:
- - 文档映射:references/openclaw-doc-map.md
- 安全清单:references/openclaw-security-checklist.md
- 模式矩阵:references/openclaw-mode-matrix.md
- 操作系统矩阵:references/openclaw-os-matrix.md
- 集成手册:references/openclaw-integrations-playbook.md
- 运维台账模式:references/openclaw-ops-ledger-schema.md
自动化辅助脚本:
- - scripts/planopenclawrollout.py
- scripts/validateopenclawenv.py
- scripts/updateopenclawops_ledger.py
默认运维台账路径:
- - ./openclaw-manager-operations-ledger.md(或由运维人员指定)
硬性停止规则(严禁绕过)
若满足以下任一条件,则停止并阻止部署/安装进程:
- 1. 必需的密钥配置文件验证失败。
- 安全清单中的强制性门控未全部通过。
- 回滚路径未记录且无负责人。
- 当前阶段的运维台账未更新。
- 请求公开暴露但未设置认证边界和令牌控制。
工作流程
1) 信息收集与范围锁定
收集并确认以下信息:
- - mode:local 或 hosted
- provider:local、fly、render、railway、hetzner、gcp
- os:macos、linux、windows-wsl2
- channels:telegram、discord、slack 的子集
- integrations:email、calendar 的子集
- environment:dev、staging、prod
- exposure:private 或 public
在继续之前,写入一条 scope_lock 台账记录:
bash
python3 scripts/updateopenclawops_ledger.py \
--ledger-file ./openclaw-manager-operations-ledger.md \
--event scope_lock \
--operator codex \
--mode hosted \
--provider fly \
--os linux \
--environment prod \
--secrets-profile hosted-fly \
--channels telegram,slack \
--integrations email,calendar \
--security-status pending \
--rollback-tested no \
--blocking-issues none \
--next-owner operator \
--next-action-date 2026-02-20
2) 生成决策完备的计划
始终先生成计划:
bash
python3 scripts/planopenclawrollout.py \
--mode hosted \
--provider fly \
--os linux \
--channels telegram,slack \
--integrations email,calendar \
--environment prod \
--exposure public \
--ledger-file ./openclaw-manager-operations-ledger.md \
--output /tmp/openclaw-rollout.md
计划输出即为执行合同。不得跳过任何部分。
3) 在任何基础设施变更前验证密钥和配置
使用配置文件感知门控验证环境:
bash
python3 scripts/validateopenclawenv.py \
--env-file .env \
--profile hosted-fly \
--json
验证强制检查项:
- - 配置文件所需的密钥
- 所需的提供商/模型替代方案
- 格式错误和重复的环境变量键
- 占位符值
- 弱网关/设置令牌
- 遗留别名警告
验证后立即写入一条 predeploy_validation 台账记录。
4) 执行模式分支
分支A:本地安装(mode=local)
- 1. 使用官方安装/入门文档进行本地设置。
- 应用 references/openclaw-os-matrix.md 中的操作系统特定命令。
- 验证启动、持久化路径和本地认证边界。
- 如果请求本地公开暴露,首先应用安全清单中的网关加固门控。
分支B:托管克隆+部署(mode=hosted)
- 1. 克隆选定的OpenClaw源代码仓库。
- 遵循 references/openclaw-doc-map.md 中的提供商手册。
- 在生产流量之前配置持久化存储。
- 配置入站/认证、密钥和健康检查。
- 验证运行时日志是否存在启动/认证错误和密钥泄露。
部署/安装完成后写入一条 deploy_complete 台账记录。
5) 安全配置渠道和集成
对于每个选定的渠道/集成:
- - 仅通过密钥管理器/环境变量注入凭据
- 运行最小冒烟测试
- 验证认证边界和错误日志安全性
将每个项目标记为:
- - configured(已配置)
- pending_credentials(凭据待定)
- blocked(已阻塞)
对于电子邮件/日历的具体细节,请使用 references/openclaw-integrations-playbook.md。
6) 代理+内存基线
记录并验证:
- - 内存持久化策略
- 保留期预期
- 重启和恢复行为
- 代理行为边界
使用运维基线决策更新台账。
7) 强制性安全门控
运行 references/openclaw-security-checklist.md 并生成每个门控的通过/失败结果。
如果任何强制性门控失败,则不得上线。
写入一条 security_gate 台账记录,明确列出阻塞项(如有)。
8) 交接和事件就绪
生成:
- - 提供商状态摘要
- 渠道+集成矩阵
- 安全门控表
- 回滚和升级负责人
- 按风险排序的后续行动
写入 handover 台账记录。对于事件/故障排查,在事件发生时追加 incident 记录。
输出合同
始终返回:
- 1. 模式/提供商/操作系统/环境状态摘要
- 安全门控结果(硬性通过/失败)
- 渠道+集成矩阵
- 代理+内存配置摘要
- 运维台账更新确认(已写入的事件名称)
- 按风险排序的后续行动