← Back to Skills Marketplace
halfmoon82

Brand Marketing Workflow

by halfmoon82 · GitHub ↗ · v2.1.0 · MIT-0
cross-platform ⚠ suspicious
112
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install halfmoon82-brand-marketing-workflow
Description
端到端品牌营销自动化工作流。从品牌输入到营销内容生产、竞品分析、 效果评估的完整闭环。支持小红书/微博/抖音多平台内容生成。
README (SKILL.md)

Brand Marketing Workflow

功能概述

端到端品牌营销自动化工作流,将品牌输入转化为可发布的营销内容资产。

核心模块

模块 功能 输出
normalize_brand_input.py 标准化品牌输入 结构化品牌参数
workflow_orchestrator.py 工作流编排 品牌简报、内容策略
content_producer.py 内容资产生产 多平台帖子/脚本/回复
competitor_fetcher.py 竞品信号抓取 公开竞品信息
competitor_ai_analyzer.py AI 竞品分析 营销洞察报告
authorization_manager.py 授权边界管理 人机协作决策
score_content_effect.py 内容效果评分 质量评估与优化建议

快速开始

1. 配置 LLM

编辑 ~/.openclaw/openclaw.json

{
  "models": {
    "providers": {
      "kimi-coding": {
        "baseUrl": "https://api.moonshot.cn/v1",
        "apiKey": "${KIMI_API_KEY}",
        "api": "openai-completions"
      }
    }
  },
  "agents": {
    "defaults": {
      "model": {
        "primary": "kimi-coding/k2p5"
      }
    }
  }
}

2. 运行 Demo

python3 run.py --demo fashion

3. 自定义输入

python3 run.py --input my_brand.json

输入格式

{
  "brand_name": "品牌名",
  "brand_positioning": "极简高端日常穿搭",
  "brand_tone": "冷静 犀利 诗意",
  "target_audience": ["都市白领", "25-40岁"],
  "use_cases": ["日常通勤", "轻社交场景"],
  "channels": ["xiaohongshu", "weibo", "douyin"],
  "content_goals": ["品牌认知", "社区建设"],
  "brand_dos": ["诗意短文案", "干净视觉语言"],
  "brand_donts": [" aggressive promotions"],
  "competitor_scope": ["竞品A", "竞品B"],
  "kpis": ["reach", "saves", "engagement_rate"]
}

技术特性

性能优化

  • 并行执行: ThreadPoolExecutor 并行 content_producer + competitor_fetcher
  • TTL 缓存: 6小时缓存机制,减少 60% API 调用
  • 指数退避: 3次重试,2^n 退避间隔

授权管理

  • 风险分级: low/medium/high 三级阈值
  • 智能跳过: 低风险 + 公开数据 = 自动放行
  • 人工确认: 发布/支付/登录等敏感操作强制确认

边界合规

  • 仅抓取公开数据
  • 禁止绕过登录/验证码
  • 禁止自动发布
  • 禁止未经批准的支付

验证状态

  • ✅ 集成测试: 26/26 passed
  • ✅ Live Mode: fashion/tech/local 三个 demo 全部通过
  • ✅ 智能 auth 跳过生效
  • ✅ K2P5 模型调用正常

项目结构

brand-marketing-workflow/
├── run.py                      # 主入口
├── scripts/
│   ├── oc_llm_client.py       # LLM 客户端(读取用户配置)
│   ├── workflow_orchestrator.py
│   ├── content_producer.py
│   ├── competitor_fetcher.py
│   ├── competitor_ai_analyzer.py
│   ├── authorization_manager.py
│   └── integration_test.py    # 集成测试
├── templates/                  # 输出模板
├── examples/                   # 示例输入
└── evidence/                   # 验证证据

依赖

  • Python >= 3.9
  • OpenClaw >= 1.0.0
  • 可选: Brave Search API Key(竞品抓取)

许可证

MIT License - 作者: halfmoon82

Usage Guidance
This package contains substantial Python code and scripts despite being listed as 'instruction-only' and the SKILL.md asks you to edit your global OpenClaw config (~/.openclaw/openclaw.json). Before running: 1) Inspect oc_llm_client.py, gateway_client.py, and install.sh for network endpoints and any code that writes to ~/.openclaw or other system files. 2) Do not run install.sh or run.py with your real openclaw.json if it contains other provider keys — create a minimal test openclaw.json containing only a dedicated (limited) API key for testing, or run in STUB_MODE where supported (autoresearch.sh supports STUB_MODE=1). 3) If you plan to use competitor fetch, only provide the optional Brave Search key if comfortable; the skill also uses external fetching (jina/r.jina.ai in cached evidence). 4) Run the demos in a sandbox/container or with network traffic monitored, and read the output logs to confirm no unexpected outbound endpoints. 5) If you need this skill for production, ask the author to update registry metadata to declare required credentials (KIMI_API_KEY, optional BRAVE_API_KEY) and provide a clear install script and audit notes. These steps reduce the risk of accidental exposure of other keys in your OpenClaw configuration.
Capability Assessment
Purpose & Capability
The code modules (content_producer, competitor_fetcher, oc_llm_client, authorization_manager, etc.) are coherent with the marketing/competitor-analysis purpose. However the registry metadata claims 'required env vars: none' and 'instruction-only' while the package contains many Python scripts and an install.sh; SKILL.md also shows a placeholder ${KIMI_API_KEY} and optional Brave Search key. The omission of those required credentials from metadata is inconsistent.
Instruction Scope
SKILL.md instructs editing the agent config (~/.openclaw/openclaw.json) to add an LLM provider and API key, and to run run.py. Reading/editing ~/.openclaw/openclaw.json can expose other provider credentials stored there; the skill's instructions do not warn about limiting which keys are present. The README claims 'only public data' and 'no automatic publishing', which is good, but the runtime code (oc_llm_client.py and gateway_client.py) is present and will read user config and contact external LLM endpoints — this is expected but should be treated as sensitive because it uses the user's model config file.
Install Mechanism
Registry says 'No install spec — instruction-only', but the bundle contains an install.sh and 17+ scripts. There is no remote download/install from unknown URLs shown in the manifest, which lowers risk, but absence of a declared install step is inconsistent with included install.sh and code. Review install.sh before running; don't run unreviewed scripts as root or in your real environment.
Credentials
The skill metadata does not declare any required env vars, yet SKILL.md demonstrates use of ${KIMI_API_KEY} and mentions an optional Brave Search API Key. The code will read ~/.openclaw/openclaw.json to discover LLM providers — that file can contain multiple API keys for different providers. Asking to edit/read that config is reasonable for an LLM-driven skill, but the metadata should explicitly declare the primary credential(s). Because the skill will read a central agent config file, it has potential access to unrelated credentials stored there; treat that as a privilege and limit what you store in openclaw.json prior to running.
Persistence & Privilege
always:false and no indication the skill auto-enables itself in other skills. The SKILL.md asks you to edit your OpenClaw config manually (not to run an automatic config writer). Nothing in the provided docs claims permanent system-wide changes beyond reading/editing ~/.openclaw/openclaw.json. Still, inspect install.sh and run.py for any code that would modify other skills/configs before running.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install halfmoon82-brand-marketing-workflow
  3. After installation, invoke the skill by name or use /halfmoon82-brand-marketing-workflow
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v2.1.0
- 新增:支持端到端品牌营销自动化工作流,覆盖内容生产、竞品分析与效果评估 - 支持多平台内容生成:小红书、微博、抖音 - 性能优化:并行执行、TTL 缓存、指数退避机制 - 增强授权与合规管理:风险分级、自动跳过、敏感操作强制确认 - 提供详细快速开始与输入格式说明 - 集成测试全部通过,项目结构与依赖更新
Metadata
Slug halfmoon82-brand-marketing-workflow
Version 2.1.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is Brand Marketing Workflow?

端到端品牌营销自动化工作流。从品牌输入到营销内容生产、竞品分析、 效果评估的完整闭环。支持小红书/微博/抖音多平台内容生成。 It is an AI Agent Skill for Claude Code / OpenClaw, with 112 downloads so far.

How do I install Brand Marketing Workflow?

Run "/install halfmoon82-brand-marketing-workflow" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Brand Marketing Workflow free?

Yes, Brand Marketing Workflow is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Brand Marketing Workflow support?

Brand Marketing Workflow is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Brand Marketing Workflow?

It is built and maintained by halfmoon82 (@halfmoon82); the current version is v2.1.0.

💬 Comments