One Draft → Five Platforms

    1

    Turns one long-form draft into platform-native versions for WeChat, Zhihu, CSDN, Xiaohongshu and Baijiahao — rewritten per reader expectation, not trimmed.

    Free

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+22 more

    One Draft → Five Platforms

    Example session with this skill installed

    Adapt this for Zhihu and CSDN. Source is a 2,800-word WeChat article about an AI customer-service rollout.

    Key facts: a store connected a general-purpose LLM directly; measured error rate was 41% over two months of chat logs. After we added 120 business Q&As — I spent 4 hours going through them one by one with the shop owner — the error rate fell to 9% and support headcount went from 3 people to 1. The article ends with a "关注我们" promo and a QR code.

    • Read your context and instructions
    • Compiled the one draft five

    Step 1 — Skeleton card (extracted before any rewriting)

    【核心结论】AI 客服上线的最大失败原因不是模型不行，是没有把企业的实际业务知识喂进去。
【支撑证据】
  - 某门店：接入通用大模型客服，答错率 41%（该店两个月聊天记录统计）
  - 补入 120 条业务问答后，答错率降到 9%
  - 客服人力从 3 人降到 1 人（同样咨询量）
【独家锚点】"我陪着店主把 120 条问答一条条过了一遍，花了 4 个小时"
【可复用片段】"哪些问题该交给 AI、哪些必须转人工"的判断清单
【必须删掉】公众号末尾的"关注我们"、二维码提示
【源文字数】2800 字

    Note what did

    not change: the conclusion, all three numbers, the four-hour detail. That is the skeleton. Everything below it is skin.

    Step 2 — Zhihu version

    The Zhihu reader's question is "how did you actually solve it", so the conclusion moves to line 1 and the background build-up dies.

    Title options

    1. 为什么很多企业的 AI 客服上线三个月就被关掉了？
    2. 接入 AI 客服后答错率反而升到 41%，我复盘出的三个原因
    3. AI 客服不是模型问题，是知识问题

    Opening, as delivered

    先说结论：AI 客服上线失败，九成不是模型不行，是没把业务知识喂进去。

    我去年帮一家门店做客服改造，第一版直接接了通用大模型。上线一周，店主截图给我看——客户问"你们家能不能开发票"，AI 答的是"发票是财务凭证，通常由税务机关管理"。

    我们统计了两个月聊天记录：答错率 41%。

    Step 3 — CSDN version

    The CSDN reader is asking "can I copy your code", so the same facts are re-cut into 现象 / 原因 / 解决步骤, with the failure shown as a literal transcript they can recognise in their own logs.

    现象

    通用大模型直接接入企业客服场景，实测答错率 41%（两个月聊天记录统计）。

    典型错误：

    客户：你们家能开发票吗？
AI：发票是财务凭证，通常由税务机关管理。

    原因

    模型没有该门店的任何业务事实，只能回答"关于税务的一般性常识"。

    解决步骤

    整理 120 条业务问答重新注入 —— 答错率 41% → 9%，客服 3 人 → 1 人。

    What actually changed

    ElementWeChat sourceZhihuCSDN
    Openingbackground build-upconclusion in line 1bare 现象 statement
    Evidence formatnarrativenumbers up frontliteral error transcript
    Section orderstory orderconclusion → cause现象 → 原因 → 解决步骤
    Ending关注我们 + QR codean open questionreproducible steps
    Promo phrasingpresentdeleteddeleted
    Facts and numbers—identicalidentical

    The three numbers (41%, 120, 9%) and the four-hour detail are identical in all three versions. That is deliberate, not lazy: those are the skeleton, and the platform is the skin.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    What it does

    One source article in. Platform-native versions out — for WeChat Official Accounts, Zhihu, CSDN, Xiaohongshu and Baijiahao, with profiles for Juejin, Weibo and Knowledge Planet (知识星球) as well.

    It does not trim, and it does not copy-paste. It rebuilds the piece around what each platform's reader came for.

    Why use it

    Every Chinese content platform has a different reader expectation, and cross-posting the same text gets punished three ways:

    • De-duplication and "low originality" flags — platforms compare against everything already published
    • Completion-rate collapse — Zhihu readers want the conclusion in the first 200 characters; a WeChat-style opening loses them
    • De-ranking — external links, promo phrasing and wrong formatting all cost you
    • Wasted inventory — a platform only gets one first impression of your account

    How it works

    1. The skill first answers one question per platform: what does this reader want to walk away with? (Zhihu: "how did you actually solve it". CSDN: "can I copy your code". Xiaohongshu: "why does this matter to me".)
    2. It extracts a skeleton card before touching the prose — core conclusion, supporting evidence, your un-fakeable first-hand anchors, and what must be deleted.
    3. It rebuilds structure, opening, evidence type and closing per platform.
    4. It ships 11-dimension platform profiles (length, structure, taboos, title rules) as reference files, plus worked before/after examples.

    What stays the same across versions

    The facts, the numbers and your first-hand detail are byte-identical in every version. Only the skin and the entrance change. Rewriting is not paraphrasing — if the numbers drift, you have made the content worse, not better.

    Who it's for

    • Chinese content operators running one piece across many platforms
    • Anyone whose "same post everywhere" strategy is producing flat numbers
    • Agents asked to "adapt this for platform X"

    Honest limitations

    • The knowledge base is Chinese platforms. It won't help with LinkedIn or Substack.
    • Platform rules change. The profiles are a strong starting point, not a guarantee against any platform's current review policy.
    • It optimizes for reader fit and platform-safety. It can't manufacture demand for a topic nobody wants.

    How to use

    Skills on Agensi run inside your AI agent. Copy the prompt, paste it into your agent, and it connects to Agensi and installs this skill for you. If a payment is needed, your agent shows it to you first.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill passes our trust checks below.

    Security scanned

    Passed our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Free forever

    No account required to browse

    Trust & safety

    Security scanned

    Verified clean 13 days ago

    • Free to download with an account

    Listed13 days ago

    What's inside

    Frequently Asked Questions