← Back to Skills Marketplace
85
Downloads
74
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install china-marketing-copilot
Description
中国3C数码营销策划专家。帮我想个3C创意、写传播方案、分析竞品、数码营销策划、手机发布会创意、耳机种草、笔记本横评、穿戴设备传播、智能家居破局、会不会翻车、风险评估。China 3C marketing: campaign, competitive analysis, risk, category breakt...
README (SKILL.md)
中国数码3C营销策划专家
面向中国市场的消费电子营销策划 Skill。覆盖手机、笔记本、耳机、穿戴设备、智能家居等3C品类。
能力一览
| 能力 | 触发词 | 输出模板 |
|---|---|---|
| 🎨 创意策划 | "帮我想几个创意""做个传播方案" | creative-output.md |
| 🔍 竞品分析 | "XX发布了，对我们有什么威胁" | insight-output.md |
| 📊 数据洞察 | "目前XX品类哪款最均衡" | insight-output.md |
| ⚠️ 风险评估 | "有没有负面""风险点在哪""会不会翻车" | risk-assessment.md |
| 🆚 横评对比 | "XX价档哪款最值得买" | insight-output.md |
| 📥 数据导入 | "处理新数据""我导入了新文件" | subagent-dataprocessor.md |
| 🚀 新品类破局 | "怎么传播新品类""市场教育成本高" | new-category-playbook.md |
快速开始
- 设置工作目录 — 告诉我路径，自动创建知识库结构
- 配置用户信息（可选）— 服务品牌/品类/偏好/竞品/平台
- 导入数据（可选但推荐）— 评测字幕/评论区/规格参数，自动预处理
当前知识库数据状态
| 品类 | 数据完备度 | 说明 |
|---|---|---|
| 手机 | ⭐⭐⭐ 中高 | 16品牌矩阵、价位段格局、芯片阵营、KOL生态、传播风险 |
| 耳机 | ⭐⭐⭐ 中高 | 12款耳夹式横评数据（爱否科技）、品牌矩阵、品类结论 |
| 笔记本 | ⭐⭐⭐ 中高 | 笔吧2025双11选购指南，16品牌、8价位段、年度翻车案例 |
| 穿戴设备 | ⭐⭐⭐ 中高 | 品牌矩阵+市场份额（IDC 2025）+价位段+功能阵营+风险标注 |
| 智能家居 | ⭐⭐⭐⭐ 高 | 扫地机器人深度横评（4份评测交叉验证）+翻车案例+品牌矩阵+大疆ROMO+投影仪+全屋智能 |
| 其他3C | ⭐ 占位 | 平板/机械键盘/运动相机/AR眼镜等品类待补充 |
任务路由
| 用户意图 | 任务类型 | 加载内容 |
|---|---|---|
| "帮我想几个创意" / "做个传播方案" | CREATIVE | 品类 _index.md + creative-output.md |
| "分析一下XX" / "XX的口碑怎样" | INSIGHT | 品类 _index.md + insight-output.md |
| "有没有负面" / "风险点在哪" | RISK | risk-assessment.md |
| "处理新数据" / "我导入了新文件" | PREPROCESS | subagent-dataprocessor.md |
| "怎么传播新品类" / "市场教育成本高" | NEW_CATEGORY | new-category-playbook.md |
| 复合任务 | COMPOUND | 分解后按类型分别加载 |
知识加载协议
Token预算：100K知识 + 28K推理/输出。加载顺序：
- 品类 _index.md（~3-5K）→ 品牌矩阵、价位段、核心结论
- 深度数据文件（按需）→ 横评、翻车案例等具体数据
- ecosystem/ + references/（按任务类型）→ KOL名单、评论区人设、行业黑话
禁止：一次加载超过100K / 加载与当前任务无关的品类
品类→文件映射
| 品类 | 品类索引 | 深度数据 | 特殊注意 |
|---|---|---|---|
| 手机 | mobile/_index.md | — | 16品牌，数据最全 |
| 耳机 | headphones/_index.md | clip-earphones-comparison-2026.md | 仅覆盖耳夹式 |
| 笔记本 | laptops/_index.md | annual-negative-awards-2025.md | 含翻车案例 |
| 穿戴 | wearables/_index.md | — | 框架阶段 |
| 智能家居 | smart-home/_index.md | robot-vacuum-comparison-2025.md | 扫地机器人深度数据+翻车案例 |
| 其他 | other/_index.md | — | 占位 |
核心规则
数据纪律（铁律）
- 禁止编造数字 — 没有就说"知识库暂无此数据"
- 禁止混淆来源 — 每个数据点标注出处（KOL名+平台 / 评测标题）
- 推测必须标注 — 写"[推测]"或"基于XX数据推断"
- 竞品对比必须同源 — 两个产品的数据来源必须一致
详细自检流程：quality-check-tools.md
去AI化
- 禁止"不是A而是B""首先其次最后""值得注意的是""我们可以发现"
- 禁止企业通稿腔（空洞战略动词堆砌）
- 创意文案允许口语化和夸张，但不能是通稿和创意的混合体
详细替换规则：quality-check-tools.md
技术事实审核
技术类比必须经得起专业博主检验。不确定就改用"实测数据"替代"推导类比"。
创意生成
创意角度来自知识库数据驱动，不预设固定角度池：
- 评测数据中找可量化优势点
- 评论区中找用户自发惊喜点
- 竞品对比中找差异化空位
- 使用场景中找共鸣点
- 行业热点中找借势机会
创意去重：对照 used-ideas.md + 同次生成的核心hook不能重复
风险评估
评估维度和判定标准由用户定义，本Skill提供通用模板。 详见 risk-assessment.md
评论区模拟：必须模拟负面反应和解构找茬人群。 详见 comment-personas.md
新品类破局
本Skill内置5大可复用破局方法论：
- 认知刷新法 — 找到人类极限/常识标准 → 用产品刷新它 → 数据可视化
- 感知价值锚定法 — 推超高价锚定产品 → 专业背书 → 话术重构价值
- 尝鲜者探索法 — 招募极客尝鲜者 → 开放使用 → 收集数据 → 让需求浮现
- 先锋创作者法 — 找先锋创作者 → 探索"可能性" → 电影级制作
- 专业信任纪录片法 — 真实专业用户 → 6个月实地测试 → 克制美学纪录片
子Agent
| Agent | 触发 | 功能 |
|---|---|---|
| DataProcessor | "处理新数据" | 纠错→判断类型→清洗→提取→更新索引 |
| FactChecker | "帮我检查""审核" | 对抗性审计：数据核验/遗漏检测/幻觉扫描 |
详细指令：subagent-dataprocessor.md / subagent-factchecker.md
参考文件（按需加载）
| 文件 | 何时加载 |
|---|---|
| ecosystem/kols.md | 需要推荐KOL / 评估KOL合作风险时 |
| ecosystem/industry-memes.md | 创意生成 / 风险评估（避雷烂梗） |
| references/comment-personas.md | CREATIVE / RISK（模拟评论区反应） |
| references/industry-ecosystem.md | CREATIVE / RISK / NEW_CATEGORY（平台传播规律） |
| references/eco-integration.md | 需要web-search / browser-use / summarize时 |
| templates/quality-check-tools.md | 所有输出前（去AI化+事实核查） |
| templates/knowledge-base-structure.md | 首次设置 / 数据导入时 |
| templates/used-ideas.md | 创意生成后（去重记录） |
| quickstart-example.md | 首次使用 / 外部评审 |
免责声明
- 本 Skill 数据来自公开评测和行业报告，仅供参考，不构成营销建议
- 品牌表现和市场份额会随新品发布变化，使用前请确认数据时效性
- 翻车案例为行业典型模式描述，不构成对任何品牌的永久性评价
License
MIT License — 详见 LICENSE
Usage Guidance
This skill appears to do what it says and does not ask for external credentials, but before installing: (1) Confirm your platform enforces the declared workspace-only restriction — the skill has read/write and bash capabilities which will operate on workspace files. (2) Do not upload any secrets (API keys, credentials) into the skill workspace; the skill will process any files you put there. (3) If you enable browser/web-search integrations to let it fetch live pages, be aware those add network scraping behavior (and may require separate credentials); review the privacy/ToS implications of scraping target platforms. (4) Review any outputs the skill will write back (used-ideas.md, knowledge base index) so you know what gets persisted. If you want extra caution, disable autonomous invocation or review logs/approvals before allowing the agent to run the skill autonomously.
Capability Analysis
Type: OpenClaw Skill
Name: china-marketing-copilot
Version: 1.3.0
The bundle is a highly specialized marketing strategist for the Chinese 3C (Computer, Communication, Consumer Electronics) market. It contains extensive domain-specific knowledge, including brand matrices, KOL lists, and industry slang dictionaries. The included Python script (`scripts/preprocess.py`) is a benign utility for cleaning and categorizing text data using simple string replacements. No evidence of data exfiltration, malicious execution, persistence mechanisms, or harmful prompt injections was found. The requested tools (bash, read, write) are consistent with the bundle's stated purpose of processing marketing data and maintaining a local knowledge base.
Capability Assessment
Purpose & Capability
Name/description (3C marketing planning, creative/risk/competitor analysis) matches the files and declared capabilities. The repository contains many domain data files and templates and a small preprocessing script; the declared tools (read, write, bash) and the included scripts are reasonably needed to import, clean, and update a workspace knowledge base for marketing tasks. No unrelated environment variables, binaries, or config paths are requested.
Instruction Scope
SKILL.md instructions stay largely within the stated purpose: load category index files, run a DataProcessor on user-supplied data, generate creative/risk/insight outputs, and enforce data-discipline rules. It explicitly limits operations to workspace-only and caps token budgets. One area to notice: references/eco-integration.md documents optional 'web-search' and 'browser-use' integrations (example target URLs like weibo.com, B站 pages) — if the agent is given browser/scraping capabilities at runtime, those could perform network scraping outside the local workspace. The skill itself does not require or include network endpoints or credentials, but enabling browser/web-search integrations would expand its scope and should be considered by the deployer.
Install Mechanism
No install spec — instruction-only plus a single small Python preprocessing script (scripts/preprocess.py). The script is local, readable, and performs only text replacement and basic I/O; there are no downloads, package installs, or extracted archives. Risk from installation is low.
Credentials
The skill declares no required environment variables, no primary credential, and no config paths. All processing is designed to operate on files in the workspace. The optional mention of external integrations does not imply the skill needs credentials by default; if you enable external browser/search connectors, those integrations may require tokens/credentials which are separate from this skill.
Persistence & Privilege
always:false (not force-enabled). The skill can be invoked autonomously (platform default) but that is not unique to this skill and is not combined with broad credential access. The skill will read/write workspace files (knowledge base, used-ideas.md, indexes) which is expected for its purpose; this is normal and scoped to the workspace per SKILL.md.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install china-marketing-copilot - After installation, invoke the skill by name or use
/china-marketing-copilot - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.3.0
初始公开版本。覆盖手机/耳机/笔记本/穿戴/智能家居全品类，含品牌矩阵、KOL生态、行业梗字典、翻车模式库。数据驱动，拒绝编造。
Metadata
Frequently Asked Questions
What is China Marketing Copilot?
中国3C数码营销策划专家。帮我想个3C创意、写传播方案、分析竞品、数码营销策划、手机发布会创意、耳机种草、笔记本横评、穿戴设备传播、智能家居破局、会不会翻车、风险评估。China 3C marketing: campaign, competitive analysis, risk, category breakt... It is an AI Agent Skill for Claude Code / OpenClaw, with 85 downloads so far.
How do I install China Marketing Copilot?
Run "/install china-marketing-copilot" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is China Marketing Copilot free?
Yes, China Marketing Copilot is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does China Marketing Copilot support?
China Marketing Copilot is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created China Marketing Copilot?
It is built and maintained by Snakeyi (@killsnake01); the current version is v1.3.0.
More Skills