← Back to Skills Marketplace
sirkayzh

Content Forge

by Kyle Zheng · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
35
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install content-forge-chinese
Description
把公众号/小宇宙播客/知识星球/财新/视频号/B站等中文内容一键丢进,生成播客、PPT、思维导图、深度报告。中文用户深度定制,预置 5 大杀手场景模板,与 IMA 笔记、飞书文档无缝联动。本地 LLM 兜底,保证可用性。Use when 用户说「丢进 NotebookLM」「转成播客」「公众号转播客」「PDF 做...
README (SKILL.md)

内容工坊

把任何中文内容丢进,生成播客 / PPT / 思维导图 / 深度分析。

本地化适配

  • 🗣 中文优先:生成结果默认中文,适合中文内容创作者
  • 📱 飞书联动:输出直接同步至飞书文档、多维表格、云盘
  • 🗄 IMA 知识库:内容自动存入 IMA 笔记/知识库,支持 AI 问答检索
  • 💻 本地 LLM 兜底:NotebookLM 不可用时,自动降级调用本地大模型

触发词

「丢进 NotebookLM」「生成 NotebookLM」「转成播客」「做成 PPT」「做成思维导图」「深度分析这篇」「转 IMA 笔记 / 飞书」

5 大场景路由(命中后读对应文档)

场景 输入 文档
公众号深度分析 → IMA / 飞书 公众号 URL scenarios/01-wechat-to-ima.md
小宇宙播客 → 飞书 / IMA 摘要 小宇宙 URL scenarios/02-xiaoyuzhou-to-feishu.md
知识星球精华 → 思维导图 帖子文本 scenarios/03-zsxq-to-mindmap.md
财新文章 → 通勤播客 财新 URL scenarios/04-caixin-to-podcast.md
视频号 / B站 → 团队 PPT 视频链接 scenarios/05-shipinhao-to-ppt.md

工作流(所有场景统一)

1. 抓取 → 2. 清洗 → 3. 【优先】NotebookLM 生成
                      ↘【降级】本地 LLM(C 方案)→ 4. 落地(IMA / 飞书 / 本地)

NotebookLM 可用性检测:每次执行 Step 3 前先运行 notebooklm status,可用则走 NotebookLM路径;失败则自动降级到 C 方案(本地 LLM 直接生成),不影响产出。

⚠️ 2026-05-28 重要更新:CLI notebooklm 全功能不可用(generate HTTP 400、source add-text 假成功),上传+生成全部改用 Playwright 浏览器自动化

  • CLI 仅用于:login(获取 cookies)/ list / create
  • 上传改用:scripts/notebooklm-playwright.js upload-text "\x3Cnotebook标题>" "\x3Csource标题>" "\x3Ccontent>"
  • 生成改用:scripts/notebooklm-playwright.js generate "\x3Cnotebook标题>" \x3Caudio\|video\|report\|mindmap\|flashcards\|quiz\|presentation\|infographic\|datatable>
  • 脚本通过 Playwright 操控浏览器完成全流程

📋 配额规则(2026-05-28 实测)

维度 类型 重置条件
🔴 账号级每日上限 音频、视频、报告 新 notebook 也无法重置(但 24h 滚动窗口后可用)
🟢 notebook 级每日 1 次 思维导图、闪卡、测验、演示文稿、信息图、数据表格 新建 notebook 即重置

匹配场景 → 读 scenarios/0X-xxx.md → 按 SOP 执行 → 失败走 references/troubleshooting.md 降级。

🛡️ 诚实度规范(所有场景必守)

⚠️ 本 skill 的核心差异化:不只追求"做出来",更追求"如实标注信息来源"

三类内容必须区分

类型 标记方式 例子
事实 直陈 节目时长 78 分钟
引用 用「」 「我们真是最后一代古代人」
AI 推断 (推断)(基于章节标题推测) "本期高潮是 AI 与时代位置的讨论 (推断)"
AI 建议 标注「推荐讲法」「AI 建议」 "可以这样开场:…… (AI 建议)"

输出底部必带 confidence 字段

{
  "source_completeness": "full | partial | metadata-only",
  "ai_inference_ratio": "low | medium | high",
  "warnings": ["未听原音频", "正文仅有节选"]
}

抓取不完整时显著警告

当只拿到 metadata 而非全文时,输出顶部 必加一行警告:

⚠️ 本分析基于 [节目页 shownotes / 章节标题 / 元数据],未读全文/未听原音频。

NotebookLM 生成内容标注

调用 NotebookLM Audio Overview / Mind Map 时,输出注明:

🤖 以下内容由 NotebookLM 基于 source 演绎生成,可能包含:
   - 未在原文出现的过渡和判断
   - AI 推测的因果关系
   原始 source: \x3Clink>

📖 详细规范 → references/honesty-rules.md

按需查阅

何时 读什么
调用任何工具前 references/tool-map.md 🔴 必读
写输出前 references/honesty-rules.md 🔴 必读
首次安装 references/installation-cn.md
网络不通 references/china-network.md
抓不到内容 references/paywall-strategies-cn.md
任意环节报错 references/troubleshooting.md
PPT 生成问题 references/troubleshooting.md →「PPT 生成工具链」章节

协同

WebFetch(内置抓取)/ lark-master(飞书)/ ima-skill:notes:knowledge-base(IMA,注意区别!)/ feishu-md-cleaner(格式清理)/ agent-browser(兜底浏览器)

⚠️ 每个工具的能力/触发/坑请查 references/tool-map.md,不要凭名字猜用法。

License & Credits

MIT · 原创:SirKayZh/ContentForge · 致谢上游:joeseesun/qiaomu-anything-to-notebooklm

Usage Guidance
Install only if you are comfortable with a skill that may use browser automation, logged-in sessions or cookies, proxies, external AI services, and cloud document systems. Use it only with content you are authorized to access and transform, avoid pasting raw session cookies, review every external upload or write destination before running, and prefer local-only/manual input paths for paywalled, confidential, or personal material.
Capability Tags
requires-oauth-tokenrequires-sensitive-credentials
Capability Assessment
Purpose & Capability
The stated content-transformation purpose is coherent for NotebookLM, Feishu, IMA, transcription, and local output, but the artifacts also explicitly include paywall-circumvention paths for Caixin and overseas publications, user-cookie injection, bot/user-agent disguise, cache/mirror retrieval, and redistribution options such as NAS, RSS, and podcast app upload.
Instruction Scope
The workflow uses broad trigger phrases and can route ordinary content requests into network fetching, browser automation, NotebookLM upload/generation, Feishu or IMA writes, and local persistence. Some scenarios ask before optional landing destinations, but the artifacts do not consistently require explicit consent before external upload, credential use, or write operations.
Install Mechanism
Installation is conventional git/pip/npm setup with optional NotebookLM login and companion skills. It is disclosed, but setup guidance also suggests persistent shell environment changes for proxies and API keys without strong rollback or secret-handling guidance.
Credentials
Playwright browser automation, local shell commands, API keys, proxies, yt-dlp, ffmpeg, OCR/transcription services, NotebookLM, Feishu, and IMA are partly expected for the workflow, but they are overbroad when combined with paywall acquisition strategies and raw authenticated-session handling.
Persistence & Privilege
The skill uses NotebookLM browser storage state, may request user login cookies, creates notebooks, uploads sources, writes to Downloads, and can sync or publish outputs to Feishu, IMA, NAS, cloud storage, RSS, or podcast apps. These are high-impact side effects and retention paths with incomplete scoping and cleanup guidance.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install content-forge-chinese
  3. After installation, invoke the skill by name or use /content-forge-chinese
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
ContentForge 中文内容工坊 1.0.0 — 一站式将各类中文内容转化为播客、PPT、思维导图等 - 支持将公众号、小宇宙、知识星球、财新、B站等中文内容一键生成多种形式产物。 - 集成 NotebookLM(含 Playwright 自动化绕过 CLI 限制)与本地 LLM,保障稳定可用。 - 提供 5 大高频场景模板,自动路由至相应 SOP 文档。 - 接入飞书、IMA 笔记,实现多平台联动和内容落地。 - 严格区分事实、引用、AI 推断和建议,输出附信心分级与数据完整性警告。 - 明确操作流程、工具映射和故障处理指南,提升易用性与可靠性。
Metadata
Slug content-forge-chinese
Version 1.0.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is Content Forge?

把公众号/小宇宙播客/知识星球/财新/视频号/B站等中文内容一键丢进,生成播客、PPT、思维导图、深度报告。中文用户深度定制,预置 5 大杀手场景模板,与 IMA 笔记、飞书文档无缝联动。本地 LLM 兜底,保证可用性。Use when 用户说「丢进 NotebookLM」「转成播客」「公众号转播客」「PDF 做... It is an AI Agent Skill for Claude Code / OpenClaw, with 35 downloads so far.

How do I install Content Forge?

Run "/install content-forge-chinese" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Content Forge free?

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

Which platforms does Content Forge support?

Content Forge is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Content Forge?

It is built and maintained by Kyle Zheng (@sirkayzh); the current version is v1.0.0.

💬 Comments