自媒体内容创作大师

部署内容创作Agent团队（墨白主编+探风选题+锦书文案），平台无关的内容生产核心。使用 /content-creation 触发，交互式引导配置品牌信息并自动部署。

Security Scan

The skill's purpose (deploy a 3-person OpenClaw content-agent team) aligns with the files and CLI calls, but there are multiple mismatches between the human-facing SKILL.md and the included script/templates that deserve review before installing.

Purpose & Capability

The name/description promise to deploy a 3-person content-creation agent team and the repository contains templates and a Node.js setup script that copies templates and registers agents via the openclaw CLI — this is coherent with the stated purpose. No unrelated credentials, binaries, or external packages are requested.

Instruction Scope

SKILL.md describes an interactive, three-round user prompt flow that collects 12 fields and then writes those answers into USER.md files (template substitution). The provided scripts/setup.cjs, however, does not implement interactive prompting or placeholder substitution: it merely copies markdown templates into ~/.openclaw/workspace-content-creation/<agent> and runs 'openclaw agents add'. Templates also reference extra placeholders (e.g., current_followers, growth_target) and some bootstraps mention 5 agents while the script deploys 3 — these inconsistencies mean the README/instructions and the runnable script do not fully match.

Install Mechanism

No remote downloads or package installs. The included script is a local Node.js script that uses fs and child_process.execSync to copy files and call the local 'openclaw' CLI. There are no external URLs, no archives to extract, and no package registry installs in the skill bundle.

Credentials

The skill requests no environment variables or secrets. The script respects an optional OPENCLAW_HOME environment variable (falling back to ~/.openclaw) — reasonable for an OpenClaw-integrating tool. It does create files under the user's home directory (~/.openclaw), which is expected for agent workspaces but worth noting.

Persistence & Privilege

The skill is not always-enabled and does not request elevated privileges. It will create a workspace directory under the user's OpenClaw home and register agents via the openclaw CLI (persistent changes within the OpenClaw ecosystem). This behavior is consistent with the stated purpose but will modify ~/.openclaw and OpenClaw's agent registry.

What to consider before installing

What to consider before installing: - Functional mismatch: SKILL.md promises interactive collection of user answers and placeholder substitution into USER.md files, but scripts/setup.cjs only copies templates and runs 'openclaw agents add' — it does not prompt the user or replace placeholders. If you expect the interactive flow, you'll need to run the prompts elsewhere or modify the script to perform replacements. - Template inconsistencies: Some templates include placeholders not collected by the SKILL.md prompts (e.g., current_followers, growth_target). Some BOOTSTRAP.md mentions a 5-person team although the script deploys 3 agents. Review and edit templates to avoid incomplete or misleading files. - File and CLI changes: The script will create ~/.openclaw/workspace-content-creation and call the local 'openclaw' CLI to register agents. Back up any existing ~/.openclaw state before running. Confirm that the 'openclaw' binary you have is the expected, trusted CLI (the script executes shell commands via execSync). - No obvious exfiltration or remote download: There are no external URLs or hidden network endpoints in the bundle, and it doesn't request secrets. However, the deployed agents/templates instruct future agent instances to perform web searches and to write memory files under the workspace; consider whether you are comfortable with those agents having read/write access to files under ~/.openclaw. - Safe testing recommendations: Run the script in a controlled environment first (non-production account or temporary user), inspect the copied template files before letting any agent run autonomously, and consider manually performing the interactive prompt-and-template-substitution step rather than relying on the provided script. If you want the full interactive behavior, update or wrap scripts/setup.cjs to implement the prompt+templating logic or rely on the platform's agent runtime to collect and write the USER.md fields as described. If you want, I can point out exact lines in scripts/setup.cjs to change to implement interactive prompting and placeholder substitution, or produce a safe 'dry-run' variant that only prints actions without writing files or registering agents.

内容创作团队 - 自动部署

当用户调用 /content-creation 时，执行以下步骤部署 3 人内容创作 Agent 团队。

概述

部署平台无关的内容创作核心团队：

• 墨白（主编/审核）、探风（选题策划）、锦书（文案创作）

Step 1：环境检查

1. 执行 which openclaw，确认 OpenClaw 已安装
   • 未安装 → 输出"请先安装 OpenClaw：https://openclaw.ai"→ 终止
2. 执行 openclaw agents list，检查是否已存在 mobai/tanfeng/jinshu
   • 已存在 → 提示用户：团队已部署，是否覆盖？用户拒绝 → 终止

Step 2：交互式收集品牌与创作信息

分三轮收集，每轮一次性提问，减少来回次数。用户任意时刻输入"取消" → 终止部署。

第一轮：账号基础（一次性提问）

向用户发送以下消息，等待回复：

我需要了解你的账号基础信息，请一次性回答以下问题（可以用序号对应）：

1. 账号/品牌名称是什么？
2. 一句话描述账号定位（是什么、为谁服务、提供什么价值）？
3. 目标读者是谁？（年龄、职业、兴趣、痛点）
4. 主打内容领域？（可多选，如：职场成长 / 科技数码 / 生活方式 / 知识科普）

第二轮：创作风格（一次性提问）

向用户发送以下消息，等待回复：

接下来了解你的内容创作偏好：

5. 文风偏好？（如：专业严谨 / 轻松幽默 / 温暖治愈 / 犀利观点）
6. 有没有 2-3 个你欣赏的对标账号或文章风格参考？（帮助团队理解你想要的感觉）
7. 你的读者最关心什么问题 / 最常遇到什么痛点？（这是选题的核心依据）
8. 有哪些话题或表达方式是绝对不能碰的？

第三轮：内容规划（一次性提问）

向用户发送以下消息，等待回复：

最后是内容规划信息：

9.  发布节奏？（如：每周 2 篇 / 每周 3 篇）
10. 每篇文章期望字数范围？（如：1500-2500 字）
11. 品牌调性关键词，3-5 个词概括你的内容气质？（如：专业、有温度、接地气）
12. 近期有没有想重点发力的内容方向或主题？（可选，没有可跳过）

将三轮答案整理后暂存备用。

Step 3：创建目录结构并部署文件

1. 创建团队根目录：

   mkdir -p ~/.openclaw/workspace-content-creation
   

2. 从 {baseDir}/templates/ 目录复制文件到 3 个子目录：

   ~/.openclaw/workspace-content-creation/mobai/
   ~/.openclaw/workspace-content-creation/tanfeng/
   ~/.openclaw/workspace-content-creation/jinshu/
   

3. 将 Step 2 收集的信息写入每个子目录的 USER.md，按以下映射替换占位符：
   • {{account_name}} → 第1题答案
   • {{positioning}} → 第2题答案
   • {{target_audience}} → 第3题答案
   • {{main_topics}} → 第4题答案
   • {{writing_style}} → 第5题答案
   • {{reference_accounts}} → 第6题答案（未填写则填"暂无"）
   • {{reader_pain_points}} → 第7题答案
   • {{forbidden_topics}} → 第8题答案
   • {{publish_frequency}} → 第9题答案
   • {{article_length}} → 第10题答案
   • {{brand_keywords}} → 第11题答案
   • {{focus_topics}} → 第12题答案（未填写则填"暂无"）
4. 每完成一个目录，输出进度：

   [1/3] mobai（墨白 - 主编）→ 已部署
   [2/3] tanfeng（探风 - 选题策划师）→ 已部署
   [3/3] jinshu（锦书 - 文案创作师）→ 已部署
   

Step 4：注册 Agent

openclaw agents add mobai \
  --name "墨白" \
  --workspace "~/.openclaw/workspace-content-creation/mobai" \
  --description "主编/内容总监 - 内容战略与质量把控"

openclaw agents add tanfeng \
  --name "探风" \
  --workspace "~/.openclaw/workspace-content-creation/tanfeng" \
  --description "选题策划师 - 热点追踪与选题规划"

openclaw agents add jinshu \
  --name "锦书" \
  --workspace "~/.openclaw/workspace-content-creation/jinshu" \
  --description "文案创作师 - 文章撰写与标题优化"

幂等性：如果 agent 已存在，跳过并提示"已存在，跳过注册"。

Step 5：验证部署

1. 执行 openclaw agents list，确认 3 个 agent 注册成功
2. 输出部署报告：

   ✅ 内容创作团队部署完成
   ├── ✍️  墨白（主编）          → 已就绪
   ├── 🔍 探风（选题策划师）    → 已就绪
   └── 📝 锦书（文案创作师）    → 已就绪
   

Step 6：后续指引

1. 使用团队：输入 /content-pipeline 启动内容创作流水线
2. 自定义角色：编辑 ~/.openclaw/workspace-content-creation/<agent>/SOUL.md
3. 如需发布到微信公众号：安装 /wechat-publisher 并使用 /wechat-publish-pipeline
4. 如需发布到其他平台：安装对应平台的发布 Skill

错误处理

• openclaw 命令不存在 → 提示安装链接，终止
• 目录创建失败 → 检查权限
• Agent 注册失败 → 检查是否重名（openclaw agents list）