copywriting

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Copywriting

文案写作

IS: short conversion copy (landing pages, hero, subheads, CTAs, product descriptions, onboarding strings, email subjects); product-state strings (destructive CTAs, error, success, empty, loading, permission copy); stripping AI writing tells from any copy.
IS NOT: long-form articles, posts, or anything written as a person rather than a brand (use the external
```
ghostwriter
```
skill with platform
```
blog
```
), slide or deck copy (use
```
presentation-creator
```
), API/product/reference docs (use
```
docs-writing
```
), in-session assistant talk (use
```
eli5
```
), or deciding which action exists, its scope, consequence, reversibility, or reachable states (use
```
product-design
```
; this skill writes final wording once those are decided).

Two modes, auto-detected:

Copy exists or the user pasted copy to fix: Mode B (Edit).
Nothing written yet, or the user wants something new: Mode A (Write).
Genuinely ambiguous ("improve this", no copy in scope): ask one question, then commit.

适用场景： 短转化文案（着陆页、核心标题、副标题、CTA、产品描述、引导流程文案、邮件主题）；产品状态文案（破坏性CTA、错误、成功、空状态、加载、权限提示文案）；清除任何文案中的AI生成痕迹。
不适用场景： 长篇文章、帖子，或以个人而非品牌身份撰写的内容（请使用外部
```
ghostwriter
```
技能并选择
```
blog
```
平台）；幻灯片或演示文稿文案（使用
```
presentation-creator
```
）；API/产品/参考文档（使用
```
docs-writing
```
）；会话助手话术（使用
```
eli5
```
）；或决定产品功能的存在、范围、后果、可逆性及可达状态（使用
```
product-design
```
；本技能仅在这些决策确定后撰写最终措辞）。

两种模式，自动检测：

已有文案或用户粘贴文案要求修改：模式B（编辑）。
无现有文案，或用户需要全新内容：模式A（撰写）。
需求模糊（如“优化这段内容”但未提供文案）：先提出一个问题，再执行任务。

Reference files

参考文件

File	Read when
`references/frameworks.md`	Pick a framework by awareness stage (Write Step 4); audit against the frameworks (Edit Step 3)
`references/page-types.md`	Copy norms for a homepage, landing, pricing, feature, or about page (Write Step 4)
`references/word-lists.md`	Flag Tier 1/2/3 AI vocabulary (Edit Step 4)
`references/ai-patterns.md`	Flag structural, sentence-level, and drafting AI tells; P0/P1/P2 triage (Edit Step 4)
`references/sweeps.md`	Run the seven line-level sweeps, then the hyphenation pass (Edit Step 5)
`references/ui-states.md`	The copy is a product state or action label, not marketing (Write Step 4; Edit Step 6 before using `[STATE-COPY]` )
`references/voice-chart.md`	No usable voice file exists and the product needs one (Write Step 3; Edit Step 1); also holds the tone-by-reader-state table
`evals/evals.json`	Only when changing this skill: behavioural scenarios with assertions plus should-trigger and near-miss routing prompts. Never loads during a user task

文件	读取时机
`references/frameworks.md`	撰写步骤4：根据用户认知阶段选择框架；编辑步骤3：对照框架审核文案
`references/page-types.md`	撰写步骤4：首页、着陆页、定价页、功能页或关于页的文案规范
`references/word-lists.md`	编辑步骤4：标记1/2/3级AI词汇
`references/ai-patterns.md`	编辑步骤4：标记结构、句子层面及草稿阶段的AI生成痕迹；按P0/P1/P2优先级分类
`references/sweeps.md`	编辑步骤5：执行7项逐行检查，最后进行连字符格式修正
`references/ui-states.md`	文案为产品状态或动作标签而非营销内容时（撰写步骤4；编辑步骤6，使用 `[STATE-COPY]` 标签前）
`references/voice-chart.md`	无可用语调文件且产品需要时（撰写步骤3；编辑步骤1）；同时包含不同用户状态对应的语调表
`evals/evals.json`	仅修改本技能时使用：包含行为场景、断言以及触发/非触发路由提示。用户任务执行期间绝不加载

Banned words

禁用词汇

The never-write set. Applies in both modes and to every line you hand back:

delve, leverage (verb), robust, seamless, holistic, paradigm, game-changing, cutting-edge, innovative, synergy, revolutionary, effortless, world-class, powerful, showcase, unlock

Also ban "simple" as a claim ("our simple onboarding"): never earned upfront, reads as an unkept promise.

Em dashes count as a banned word: none in headings or body, and no

--

or spaced hyphen standing in for one. Use a comma, colon, full stop, or parentheses.

A voice file that names one of these as a signature word is the only thing that overrides the list (Write Step 3).

references/word-lists.md

holds the wider tiered vocabulary with replacements; that list is for detection in Edit mode, not a second copy of this one.

绝对禁用词汇。两种模式下返回的所有内容均适用：

delve, leverage（动词形式）, robust, seamless, holistic, paradigm, game-changing, cutting-edge, innovative, synergy, revolutionary, effortless, world-class, powerful, showcase, unlock

另外，禁止将**"simple"**用作宣传语（如"our simple onboarding"）：该词汇无法预先验证，会给人一种未兑现承诺的感觉。

破折号也属于禁用范畴：标题和正文中均不得使用，也不能用

--

或带空格的连字符替代。请使用逗号、冒号、句号或括号。

只有当语调文件明确将某个禁用词汇列为品牌标志性词汇时，才可例外（撰写步骤3）。

references/word-lists.md

包含更全面的分级词汇及替代词；该列表仅用于编辑模式下的检测，并非本禁用列表的重复版本。

Mode A: Writing new copy

模式A：撰写新文案

Writing progress:
- [ ] Step 1: Gather context
- [ ] Step 2: State the brief, then write
- [ ] Step 3: Discover brand voice
- [ ] Step 4: Route by copy type, choose framework
- [ ] Step 5: Write 2-3 alternatives
- [ ] Step 6: Recommend and explain
- [ ] Step 7: Check every option against the brief

写作进度：
- [ ] 步骤1：收集上下文
- [ ] 步骤2：明确需求简报，开始撰写
- [ ] 步骤3：确定品牌语调
- [ ] 步骤4：按文案类型分类，选择框架
- [ ] 步骤5：撰写2-3种备选方案
- [ ] 步骤6：推荐并说明理由
- [ ] 步骤7：对照需求简报检查所有方案

Step 1: Gather context

步骤1：收集上下文

Settle four things, from the user or from the files. Where the files do not settle one, infer it and name the inference in Step 2; the failure mode is an invented audience or goal presented as fact.

Page purpose. The one action this page drives (sign up, book a demo, download).
Audience. The specific reader: job title, pain, what they've already tried.
Product. What it does; the concrete user outcome.
Traffic source. Where the reader comes from (cold ad, warm email, organic search, referral, inside the product).

Traffic source sets the reader's awareness stage (

references/frameworks.md

): cold readers need the problem named before the product; warm readers already know the problem and want the mechanism or proof.

从用户或参考文件中确定以下四点。若参考文件未明确说明，则在步骤2中注明推断内容；需避免将虚构的受众或目标当作事实。

页面目标：该页面要引导用户完成的核心动作（注册、预约演示、下载）。
受众：具体读者群体：职位、痛点、已尝试过的解决方案。
产品：产品功能；能为用户带来的具体成果。
流量来源：读者的访问渠道（冷启动广告、暖邮件、自然搜索、推荐、产品内部）。

流量来源决定了读者的认知阶段（

references/frameworks.md

）：冷启动读者需要先了解问题，再介绍产品；暖启动读者已了解问题，想知道解决方案或证明。

Step 2: State the brief, then write

步骤2：明确需求简报，开始撰写

State the brief and keep going. Mark every field you inferred rather than were told, so the user corrects it against real copy instead of against a question:

Brief:
- Page: [page type]
- Goal: [single action]
- Reader: [specific audience]
- Core outcome: [what changes for the reader]
- Tone: [inferred from brand voice or user-stated]
- Awareness: [unaware / problem-aware / solution-aware / product-aware / most-aware]

Inferred (correct me): [fields you guessed]

Stop and ask before writing only when a wrong guess makes the work useless or unsafe: the copy ships in this turn with no review, or the goal is genuinely unknown and each candidate goal produces different copy.

明确需求简报后继续撰写。标注所有推断得出的信息，以便用户对照实际文案进行修正，而非针对问题回答：

需求简报：
- 页面类型：[页面类型]
- 核心目标：[单一动作]
- 读者群体：[具体受众]
- 核心成果：[读者能获得的改变]
- 语调：[从品牌语调或用户需求推断]
- 认知阶段：[未知 / 问题感知 / 解决方案感知 / 产品感知 / 完全了解]

推断内容（如有误请修正）：[你猜测的信息]

仅当错误推断会导致工作无效或存在风险时才停止撰写并提问：比如文案将直接发布无需审核，或目标完全未知且不同目标会产生截然不同的文案。

Step 3: Discover brand voice

步骤3：确定品牌语调

Find voice signals before inventing one. Work down this order and stop at the first hit:

A voice file in the repo:
```
VOICE.md
```
,
```
BRAND.md
```
,
```
docs/voice.md
```
, or a tone-of-voice doc. Authoritative when it exists.
The user's own voice, on request only. When the user asks for their voice ("in my voice", "sound like me") and the brand is theirs, read
```
$GHOSTWRITER_HOME/soul.md
```
, falling back to
```
~/.config/ghostwriter/soul.md
```
. A personal voice belongs only to the person's own brand, and the file's contents stay out of the output, the brief, and the repo.
Existing copy: copy files, README headers, or shipped marketing pages.
Inference: B2B SaaS direct and confident, consumer apps warmer, developer tools terse and honest. Ask for brand guidelines alongside the draft, not instead of it.

A discovered voice outranks the word lists: a listed word the voice file or shipped copy uses as a signature stays. Locale and spelling convention come from the voice too. Note in the brief which source you used, and mark the voice as inferred when it came from step 4. When no voice file exists and the product will need one, load

references/voice-chart.md

and offer to write

VOICE.md

alongside the copy.

Voice is constant; tone adapts to the reader's state (frustrated, confused, confident, cautious, successful). The table in

references/voice-chart.md

sets the expected shifts; copy that keeps one register across all five reads as robotic in the good moments and cold in the bad ones.

先寻找语调线索，再自行创造。按以下顺序查找，找到第一个有效线索后停止：

仓库中的语调文件：
```
VOICE.md
```
、
```
BRAND.md
```
、
```
docs/voice.md
```
或语调指南文档。若存在则具有权威性。
用户个人语调（仅按需使用）：当用户要求使用其个人语调（如“用我的风格”、“像我一样表达”）且品牌属于用户个人时，读取
```
$GHOSTWRITER_HOME/soul.md
```
，若不存在则 fallback 到
```
~/.config/ghostwriter/soul.md
```
。个人语调仅适用于用户个人品牌，文件内容不得出现在输出、简报或仓库中。
现有文案：文案文件、README标题或已发布的营销页面。
推断：B2B SaaS产品直接自信，消费类应用更亲切，开发者工具简洁诚实。撰写草稿时可同时请求品牌指南，而非先等待指南。

已确定的语调优先级高于词汇列表：若语调文件或已发布文案将某个禁用词汇列为标志性词汇，则保留该词汇。地区和拼写规范也从语调文件获取。在简报中注明你使用的语调来源，若来自步骤4则标记为推断。若无可用语调文件且产品需要，则加载

references/voice-chart.md

并提议同时撰写

VOICE.md

。

语调保持一致，但语气需根据用户状态调整（沮丧、困惑、自信、谨慎、成功）。

references/voice-chart.md

中的表格规定了语气变化；若所有场景使用同一语气，在愉悦场景会显得机械，在负面场景会显得冷漠。

Step 4: Route by copy type, choose framework

步骤4：按文案类型分类，选择框架

Product-state copy (error, empty, success, loading, permission) or an action label: load
```
references/ui-states.md
```
and stop here. Persuasion frameworks do not apply to a button that deletes something.
Marketing copy: load
```
references/frameworks.md
```
, plus
```
references/page-types.md
```
when the target is a homepage, landing, pricing, feature, or about page.

Choose the lead framework from the brief's awareness stage:

Reader	Lead framework
Unaware or problem-aware, cold traffic	Why/How/What, or PAS when the pain is sharp
Solution-aware, comparing options	Benefit Not Feature, backed by Show Don't Tell
Product-aware, low trust	Show Don't Tell, proof-led
Most-aware, transactional page	CTA Clarity
Long-form sales page, any stage	PAS or BAB as the spine, others layered inside

Layer frameworks freely. Hero copy almost always opens with Why (the reader's motivation), whatever else it uses.

产品状态文案（错误、空状态、成功、加载、权限提示）或动作标签：加载
```
references/ui-states.md
```
，流程到此为止。说服框架不适用于删除类按钮等文案。
营销文案：加载
```
references/frameworks.md
```
，若目标为首页、着陆页、定价页、功能页或关于页，同时加载
```
references/page-types.md
```
。

根据需求简报中的认知阶段选择主导框架：

读者认知阶段	主导框架
未知或问题感知，冷流量	Why/How/What框架，或痛点强烈时使用PAS框架
解决方案感知，对比选项	利益而非功能框架，辅以展示而非告知框架
产品感知，信任度低	展示而非告知框架，以证明为主
完全了解，交易页面	CTA清晰框架
长篇销售页，任何阶段	PAS或BAB框架为核心，叠加其他框架

可自由组合框架。核心标题几乎总是以Why（用户动机）开头，无论使用其他何种框架。

Step 5: Write 2-3 alternatives

步骤5：撰写2-3种备选方案

Label them Option A, Option B, Option C. Three for a page, hero, or campaign; two for a single string like a CTA or subject line, where a third is padding. One is not a choice, and four is a survey. Each option applies the chosen framework visibly, includes a headline, subhead, and at least one CTA, and is structurally different from the others, not the same idea with new adjectives.

标记为方案A、方案B、方案C。页面、核心标题或活动文案提供3种方案；单个文案（如CTA或邮件主题）提供2种方案，第三种属于冗余。仅提供1种方案没有选择空间，4种则类似调研。每个方案需明确应用所选框架，包含标题、副标题及至少一个CTA，且结构与其他方案不同，而非仅替换形容词。

Step 6: Recommend and explain

步骤6：推荐并说明理由

Pick one; state which and why in one sentence. For each unpicked option, give one specific edit note.

选择一个方案；用一句话说明选择理由。对于未选中的方案，各给出一条具体的修改建议。

Step 7: Check every option against the brief

步骤7：对照需求简报检查所有方案

Two tells survive a good draft and only show up when you hold the option against the brief: prompt echo (the headline hands the brief's own wording back) and generic default (a number, name, price, or integration the user supplied has become "thousands of teams" or "your favourite tools"). Fix both before handing back, along with any banned word. An option containing a banned word is not an option to present.

优秀草稿仍可能存在两种问题，仅当对照需求简报时才会发现：提示回声（标题直接重复需求简报的措辞）和通用默认值（用户提供的数字、名称、价格或集成信息被替换为“数千团队”或“你喜爱的工具”）。提交前修正这两个问题，以及任何禁用词汇。包含禁用词汇的方案不得提交。

Mode B: Editing existing copy

模式B：编辑现有文案

Set the edit posture before running the workflow:

Point edit: The user named one line, word, or section. Read enough surrounding copy to preserve context, change only the target plus the minimum connective tissue, and return the final wording. A point edit does not become a page audit.
Restoration: The copy already has a clear voice, angle, or opinion. Preserve its vocabulary level, relative emphasis, deliberate omissions, sentence shape, and positioning. Fix specific failures without rebalancing the argument or replacing its lead with a cleverer one.
Rebuild: The copy is generic, contradictory, or has no discernible perspective. Run the full workflow with latitude to reconstruct it from the brief, and invent no proof.

Editing progress:
- [ ] Step 1: Read all copy-bearing files
- [ ] Step 2: Set the north star
- [ ] Step 3: Audit against persuasion frameworks
- [ ] Step 4: Remove AI writing patterns
- [ ] Step 5: Run seven sweeps
- [ ] Step 6: Flag weakest elements with labels
- [ ] Step 7: Rewrite flagged sections
- [ ] Step 8: Output before/after diff

执行工作流前先确定编辑姿态：

局部编辑：用户指定某一行、某个词或某个段落。阅读足够的上下文以保持连贯性，仅修改目标内容及必要的衔接部分，返回最终措辞。局部编辑不得扩展为页面审核。
修复优化：文案已有清晰的语调、角度或观点。保留其词汇水平、相对重点、刻意省略的内容、句子结构及定位。仅修复特定问题，不得重新平衡论点或替换核心观点。
重构改写：文案通用、矛盾或无明确立场。按完整工作流执行，可根据需求简报重构，但不得虚构证明内容。

编辑进度：
- [ ] 步骤1：阅读所有含文案的文件
- [ ] 步骤2：确立核心目标
- [ ] 步骤3：对照说服框架审核
- [ ] 步骤4：移除AI生成模式
- [ ] 步骤5：执行7项逐行检查
- [ ] 步骤6：标记最弱内容并添加标签
- [ ] 步骤7：重写标记部分
- [ ] 步骤8：输出前后对比差异

Step 1: Read all copy-bearing files

步骤1：阅读所有含文案的文件

Scan every reader-facing surface: README headers, landing components, hero, CTAs, product descriptions, feature lists, onboarding strings, meta descriptions, email subjects. Read the voice file too if one exists (

VOICE.md

BRAND.md

docs/voice.md

); it settles register and locale, and it overrides the word lists for any word it names as a signature. Where the audit keeps turning on a voice question nobody has answered, load

references/voice-chart.md

and offer to settle it. Ask which files if unclear; judge copy only in the context you have read it in.

扫描所有面向读者的内容：README标题、着陆页组件、核心标题、CTA、产品描述、功能列表、引导流程文案、元描述、邮件主题。若存在语调文件（

VOICE.md

、

BRAND.md

、

docs/voice.md

）也需阅读；它确定了语域和地区规范，且若将某个词汇列为标志性词汇，可覆盖词汇列表。若审核中反复出现未解决的语调问题，加载

references/voice-chart.md

并提议解决该问题。若不确定文件范围可询问；仅根据已阅读的上下文评判文案。

Step 2: Set the north star

步骤2：确立核心目标

Write one sentence before auditing: "[User] can now [do X] without [old pain]." Every flag and rewrite serves it. If you can't write it confidently, ask; the copy is unfixable until the value proposition is clear.

审核前先写一句话：“[用户]现在可以[完成X]，无需[旧痛点]。”所有标记和改写都需服务于该目标。若无法自信地写出这句话，请询问；价值主张明确前，文案无法修复。

Step 3: Audit against persuasion frameworks

步骤3：对照说服框架审核

Load

references/frameworks.md

. Check every major copy block against each framework and carry forward only the highest-impact problems; Step 6 sets the flag budget.

加载

references/frameworks.md

。对照每个框架检查所有主要文案块，仅保留影响最大的问题；步骤6限制标记数量。

Step 4: Remove AI writing patterns

步骤4：移除AI生成模式

Load

references/word-lists.md

and

references/ai-patterns.md

. Flag each AI-ism with

[AI-ISM]

plus its pattern type:

Tier 1 words (
```
word-lists.md
```
): always flag and replace.
Tier 2 clusters (
```
word-lists.md
```
): flag when 2+ appear in one paragraph.
Structural patterns (
```
ai-patterns.md
```
): formulaic openings, chatbot artefacts, "let's" transitions, engagement hooks, rhetorical-question openers, negative parallelisms, participle tails, significance inflation, copula avoidance, em dashes as ordinary punctuation.
Drafting tells (
```
ai-patterns.md
```
section 7): prompt echo, a supplied specific swapped for a generic default, uniform confidence. These survive a word-level pass, so check them separately.

Skip for persuasion-only edits. If the user asked for AI pattern removal, run this first, before the sweeps.

加载

references/word-lists.md

和

references/ai-patterns.md

。为每个AI-isms标记

[AI-ISM]

及模式类型：

1级词汇（
```
word-lists.md
```
）：必须标记并替换。
2级词汇簇（
```
word-lists.md
```
）：同一段落出现2个及以上时标记。
结构模式（
```
ai-patterns.md
```
）：公式化开头、聊天机器人痕迹、“let's”过渡、互动钩子、反问句开头、否定排比、分词结尾、夸大重要性、避免系动词、破折号用作普通标点。
草稿痕迹（
```
ai-patterns.md
```
第7节）：提示回声、用户提供的具体信息被替换为通用默认值、过度统一的语气。这些问题无法通过词汇检查发现，需单独检查。

仅优化说服力时可跳过此步骤。若用户要求移除AI生成模式，需先执行此步骤，再进行逐行检查。

Step 5: Run seven sweeps

步骤5：执行7项逐行检查

Load

references/sweeps.md

; run all seven in order, each targeting a distinct failure mode. Finish with the compound adjective hyphenation pass at the end of that file, fixing what it catches silently rather than flagging it.

加载

references/sweeps.md

；按顺序执行所有7项检查，每项针对不同的问题类型。最后执行该文件末尾的复合形容词连字符修正，默默修复问题而非标记。

Step 6: Flag weakest elements

步骤6：标记最弱内容

Attach a label inline to every weak line. Use exactly these labels:

Label	Meaning
`[WHAT-NOT-WHY]`	Leads with product/feature, not user motivation
`[FEATURE-NOT-BENEFIT]`	Describes what the product has, not what changes for the user
`[TELL-NOT-SHOW]`	Adjective claim without proof ("powerful", "seamless", "easy")
`[VAGUE]`	Generic; could describe any product in the category
`[PASSIVE]`	Subject is acted upon instead of acting
`[VOICE-DRIFT]`	Breaks from the dominant voice of the surrounding copy (register, tense, or person)
`[PAIN-NOT-NAMED]`	States benefits without naming the frustration the reader arrived with
`[DEAD-WEIGHT]`	Adds nothing not already conveyed; safe to cut
`[JARGON]`	Technical term that obscures meaning for non-experts
`[NO-PROOF]`	Claim needing a number, example, or testimonial
`[WEAK-CTA]`	CTA describes the action, not the outcome
`[STATE-COPY]`	Vague, leaky, or dead-end state string (error, success, empty, loading, permission), or a destructive CTA labeled "Confirm"/"OK"/bare verb. Load `references/ui-states.md` before using this label; it holds the rule IDs `product-design` cites
`[AI-ISM]`	AI writing pattern: Tier 1 word, Tier 2 cluster, or structural tell

Flag the 3-7 weakest elements, prioritised by impact on conversion or comprehension. A list of twenty issues dilutes into one nobody acts on.

为每段薄弱内容添加行内标签。仅使用以下标签：

标签	含义
`[WHAT-NOT-WHY]`	以产品/功能开头，而非用户动机
`[FEATURE-NOT-BENEFIT]`	描述产品拥有的功能，而非用户能获得的改变
`[TELL-NOT-SHOW]`	仅使用形容词宣称（如“powerful”、“seamless”、“easy”），无证明
`[VAGUE]`	通用模糊；可描述同类任何产品
`[PASSIVE]`	主语为受事者而非施事者
`[VOICE-DRIFT]`	偏离周围文案的主导语调（语域、时态或人称）
`[PAIN-NOT-NAMED]`	仅说明利益，未提及读者的痛点
`[DEAD-WEIGHT]`	未添加任何新信息；可安全删除
`[JARGON]`	技术术语，对非专业人士造成理解障碍
`[NO-PROOF]`	宣称需要数字、示例或 testimonial 支持
`[WEAK-CTA]`	CTA仅描述动作，未说明成果
`[STATE-COPY]`	模糊、有漏洞或无后续的状态文案（错误、成功、空状态、加载、权限提示），或破坏性CTA标记为“Confirm”/“OK”/单一动词。使用此标签前需加载 `references/ui-states.md` ；该文件包含 `product-design` 引用的规则ID
`[AI-ISM]`	AI生成模式：1级词汇、2级词汇簇或结构痕迹

标记3-7个最弱内容，按对转化或理解的影响优先级排序。列出20个问题会稀释重点，无人会采取行动。

Step 7: Rewrite flagged sections

步骤7：重写标记部分

Cut hard: a block that reads as already-tight usually isn't. Same meaning in half the words.
Lead with Why (the user's problem or desire), not What (the product).
Name the concrete outcome, not the capability. Replace adjectives with proof: "powerful analytics" becomes "see which pages kill signups".
Make CTAs outcome-specific and short: "Start syncing" beats "Get started"; the feature explanation lives above the button, not in it.
When replacing an AI-ism, rewrite the sentence; a synonym swap moves the tell one tier down.

大胆删减：看似紧凑的文案通常仍有压缩空间，用一半字数表达相同含义。
以Why（用户的问题或需求）开头，而非What（产品）。
说明具体成果，而非功能。用证明替换形容词：“powerful analytics”改为“查看哪些页面导致注册流失”。
CTA需明确成果且简短：“Start syncing”优于“Get started”；功能说明应在按钮上方，而非按钮内。
替换AI-isms时需重写句子；仅替换同义词会让痕迹降级，仍显得像机器编辑。

Step 8: Output before/after diff

步骤8：输出前后对比差异

markdown

undefined

markdown

undefined

Copy Audit: [file or component name]

文案审核：[文件或组件名称]

North star: [one-sentence value prop]

核心目标： [一句话价值主张]

[Section name]

[章节名称]

Before:

[original text]

Issues:

[LABEL]

[LABEL]

After:

[rewritten text]

Why: [one sentence explaining the change]

修改前：

[原文内容]

问题：

[标签]

[标签]

修改后：

[改写内容]

修改理由： [一句话说明]

Summary

总结

N issues flagged across N sections
Top pattern: [most common label]
Confidence: [high / medium; note if copy context was limited]


Every "After" line keeps every fact, number, and link from its "Before", and every change fixes a named failure from the audit. A rewrite that is merely different is a regression: restore the original.

---

共标记N个问题，涉及N个章节
最常见问题：[出现次数最多的标签]
置信度：[高 / 中等；若文案上下文有限请注明]


所有“修改后”内容需保留“修改前”的所有事实、数字和链接，且每一处修改都针对审核中标记的问题。仅为不同而改写属于倒退：需恢复原文。

---

Gotchas

注意事项

"Get started" and "Learn more" attract clicks and mislead: NN/g found readers who wanted pricing or basic information landing in signup flows and losing trust. Name the destination or outcome ("See pricing", "Start a 14-day trial") instead.
The Tier 1 list dates fast. "delve" and "tapestry" were 2023-2024 tells and have largely gone; current output leans on "enhance", "highlighting", "showcasing", "align with", and participle tails ("...ensuring a smooth experience"). A pass that only greps the word table passes copy that still reads as generated; run the structural section of
```
ai-patterns.md
```
every time.
Swapping a flagged word for its neighbour in the same table ("leverage" to "harness", "robust" to "comprehensive") reintroduces the tell one tier down and reads as edited-by-machine. Rewrite the sentence around the specific instead.
```
frameworks.md
```
lists unearned adjectives (powerful, seamless, robust) that also sit in the banned words and in
```
word-lists.md
```
. One occurrence is one flag; stacking
```
[TELL-NOT-SHOW]
```
and
```
[AI-ISM]
```
on the same word inflates the count and hides the real problems.
A destructive dialog labeled "Confirm", "OK", or "Yes" forces the user to reconstruct the consequence from body text they skipped;
```
product-design
```
will cite
```
rule/no-confirm-ok-labels
```
back at you. Verb plus object ("Delete project") every time; the bare-verb exemption covers only Save, Cancel, and Close.
Template variables hide the hyphenation rule:
```
{{days}} day trial
```
ships as "7 day trial" on every plan. The hyphen goes between the variable and the unit (
```
{{days}}-day trial
```
), and never onto a standalone noun phrase ("expires in {{days}} days").
A US-spelling rewrite on an en-AU product ships as a regression across every string it touches. Read three existing strings for locale before changing one.
The personal voice lookup in Step 3 is opt-in and read-only. Reading
```
soul.md
```
does not make this the skill for personal messages, posts, or long-form; those stay with
```
ghostwriter
```
.

“Get started”和“Learn more”虽能吸引点击，但会误导用户：NN/g研究发现，想要定价或基础信息的读者进入注册流程后会失去信任。请明确目的地或成果（如“查看定价”、“开始14天试用”）。
1级词汇列表更新快。“delve”和“tapestry”是2023-2024年的AI生成痕迹，现已基本消失；当前AI输出常用“enhance”、“highlighting”、“showcasing”、“align with”及分词结尾（如“...ensuring a smooth experience”）。仅通过词汇表搜索会遗漏仍显AI生成特征的文案；每次都需执行
```
ai-patterns.md
```
中的结构检查。
将标记词汇替换为同表中的近义词（如“leverage”改为“harness”、“robust”改为“comprehensive”）会让痕迹降级，仍显得像机器编辑。需围绕具体内容重写句子。
```
frameworks.md
```
列出的无依据形容词（powerful、seamless、robust）同时出现在禁用词汇和
```
word-lists.md
```
中。出现一次只需标记一次；同一词汇同时标记
```
[TELL-NOT-SHOW]
```
和
```
[AI-ISM]
```
会夸大问题数量，掩盖真实问题。
破坏性对话框标记为“Confirm”、“OK”或“Yes”会迫使用户从跳过的正文中推断后果；
```
product-design
```
会引用
```
rule/no-confirm-ok-labels
```
规则反驳。需使用“动词+宾语”（如“Delete project”）；仅Save、Cancel和Close可使用单一动词。
模板变量会隐藏连字符规则：
```
{{days}} day trial
```
会在所有套餐中显示为“7 day trial”。连字符应放在变量和单位之间（
```
{{days}}-day trial
```
），且不得用于独立名词短语（如“expires in {{days}} days”）。
针对英式英语（澳式）产品使用美式英语改写会导致所有文案出现倒退。修改前先阅读3条现有文案确定地区规范。
步骤3中的个人语调查询为可选且只读。读取
```
soul.md
```
并不意味着本技能可用于个人消息、帖子或长篇内容；这些内容仍需使用
```
ghostwriter
```
。

Skill handoffs

技能交接

When	Run
After rewriting technical documentation copy	`docs-writing`
To optimise meta descriptions and page titles	`optimise-seo`
To review the full UI including copy in context	`ui-design` (Audit mode)
Landing page visual design, CRO strategy, conversion benchmarks	`ui-design` (Direction mode, marketing track)
The product decision of which action exists and its scope and consequence	`product-design`
Casing of headings and labels (sentence case versus title case)	`typography-audit` ( `punct-case-rules` )
In-session assistant talk, recaps, or ELI5	`eli5`

场景	调用技能
重写技术文档文案后	`docs-writing`
优化元描述和页面标题	`optimise-seo`
结合上下文审核完整UI（含文案）	`ui-design` （审核模式）
着陆页视觉设计、CRO策略、转化基准	`ui-design` （方向模式，营销赛道）
决定产品功能的存在、范围及后果	`product-design`
标题和标签的大小写（句首大写 vs 标题大小写）	`typography-audit` （ `punct-case-rules` ）
会话助手话术、总结或ELI5内容	`eli5`