知识付费流水线 · 创作技能

这个技能解决什么问题

把知识产品做成可复制的交付系统：来源整理、知识库、写作风格、公众号草稿、知识星球交付与私域跟进。

来源与使用

作者 yuanyuanyuan；许可 MIT。已核对来源与包结构，未据此宣称完成运行实测。

查看原始来源

机器可读详情 · 读取 SKILL.md · 下载完整技能包

让 Agent 使用这个技能 ↗

技能说明全文

---
name: ganhuo-ai-knowledge-commerce
description: Designs a reproducible AI-assisted knowledge-commerce pipeline for公众号、知识星球、飞书知识库、私域社群、课程陪跑、内容产品和知识付费项目. Use when the user wants to turn source research, AI drafting, style distillation, publishing, community delivery, and data review into a practical SOP that improves throughput, quality, conversion, and operating safety.
---

# 干活 AI 知识付费流水线

## Overview

Apply this skill to turn a knowledge-product idea into a repeatable operating system: source selection, source cards, AI knowledge base, writing style, content factory,公众号草稿, 知识星球交付, private-domain follow-up, and weekly review.

Keep the posture growth-oriented: improve source quality, improve content velocity, improve trust, improve conversion, and lower avoidable complaint, copyright, platform, and refund risk. Avoid treating AI rewriting as a magic cleanup layer; treat AI as the production assistant and the human operator as the editor, source owner, and final publisher.

## Minimum Input Packet

Collect or infer these fields before building the brief. If the user gives too little context, ask only for the missing fields that would materially change the output; otherwise state assumptions and produce a first draft.

- niche or category;
- target user and current pain;
- current stage: idea, first articles, existing audience, existing paid product, or optimization;
- available sources and their permission status;
- owned experience, test logs, screenshots, cases, or user questions;
- target channels: 公众号, 知识星球, 飞书知识库, 企微, 朋友圈, 小红书, or other;
- paid deliverable and rough price range;
- available tools: Feishu, n8n, RAG tool, WeChat account, analytics, CRM;
- forbidden claims, sensitive topics, and platform constraints.

## Workflow

1. Define the offer: niche, user pain, paid deliverable, price ladder, trust proof, and the reason this product is better than scattered free content.
2. Build the source engine: separate owned experience, authorized material, public references, public reports, interviews, tool tests, and user-submitted cases.
3. Create source cards: record URL or contract, author, date, license or permission status, usable scope, key points, risk notes, and what original angle will be added.
4. Build the knowledge base: organize source cards into topic clusters, user questions, templates, examples, objections, glossary, and case notes.
5. Distill the voice: define tone, pacing, examples, taboo words, title style, CTA style, and the human story bank. Improve recognizability without copying another creator's expression.
6. Run the content factory: generate topic options, outline, draft, source-linked claims, tables, CTA, and platform-specific variants.
7. Review before publishing: check source traceability, factual accuracy, personal information, commercial claims, AI-generated media disclosure, and platform fit.
8. Ship to channels: create公众号草稿, prepare知识星球主题, generate朋友圈/企微 follow-up copy, and keep public release under human approval.
9. Review the loop: record reads, completion, follows, comments, paid conversion, refund signals, questions, and next week's experiments.

## Operating Stack

Prefer the smallest stack that can be reproduced:

- Feishu Bitable: master content table, source cards, platform action table, metrics table.
- Feishu Docs or a local Markdown folder: canonical draft body and knowledge notes.
- Dify, AnythingLLM, RAGFlow, Open WebUI, LlamaIndex, or Haystack: knowledge base and RAG layer.
- n8n: schedule, RSS/API intake, status updates, review forms, and draft creation.
- WeChat Official Account API or manual editor:公众号草稿 creation, with final publish approval by a person.
- 知识星球: manual publishing and member operation unless an official API is clearly available.

Route to heavier frameworks only when the simple stack is no longer enough:

- LangGraph or CrewAI: multi-agent drafting, critique, and review loops.
- Postgres: durable audit log when Feishu Bitable is not enough.
- Custom scripts: bulk import, export, or deterministic linting.

## Output

Return a compact Knowledge Commerce Pipeline Brief:

- product positioning;
- source map and source-card schema;
- knowledge-base structure;
- publishing workflow;
-公众号 article template;
-知识星球 weekly operating rhythm;
- automation map;
- review checklist;
- 30/60/90-day rollout;
- metrics and experiment backlog.

Use tables for operating design. Limit the first execution plan to the smallest set that improves output quality or conversion this week.

## Growth Copy Rules

Transform hard guarantee language into improvement language:

- Replace "保证收入" with "提高变现概率" or "跑通收入模型".
- Replace "100% AI 写" with "AI 起草, 人工加经验和审稿".
- Replace "搬运资料" with "授权资料、公开资料导读、自有案例、实测复盘".
- Replace "自动赚钱" with "提高选题、写作、交付和复盘效率".
- Replace "必爆" with "提高完读、收藏、转发和转化的可验证动作".

Preserve energy and commercial clarity, but remove claims that cannot be proven.

## References

Read these only when needed:

- `references/checklist.md` for the brief template and scoring checklist.
- `references/platform-map.md` for Feishu, WeChat, 知识星球, and automation boundaries.
- `references/source-policy.md` for source-card rules and safer source categories.
- `references/prompt-pack.md` for reusable prompts.

## Operating Principles

- Improve the user's business goal without turning the SOP into a loophole playbook.
- Prefer one repeatable publishing loop over a giant theoretical funnel.
- Keep AI-generated output grounded in source cards and human experience.
- Keep final public publishing and refund-sensitive decisions under human control.
- Treat user trust as the compounding asset: better claims, clearer limits, faster delivery, and visible proof.

技能是任务材料，不会自动授予账户权限、付费授权或其他系统权限。请让 Agent 按说明核对依赖，再在授权范围内使用。