- Load
~/.claude/plugins/config/claude-for-legal-zh/product-legal/CLAUDE.md→ marketing claim standards. - Apply the claim taxonomy and review workflow below.
- Extract each claim. Classify: puffery/factual/comparative/implied/absolute.
- For each non-puffery claim: substantiation check, suggested revision.
- Output: item-by-item analysis with a verdict, plus suggested revised text if short enough.
/product-legal:marketing-claims-review
[粘贴落地页文案]
Matter context
Matter context. Check ## 事项工作空间in the practice-level CLAUDE.md. If Enabled is ✗(the default for enterprise legal users), skip the rest of this paragraph—the skill uses practice-level context and the matter mechanism is invisible. If enabled and there is no active matter, ask: "Which matter is this? Run /product-legal:matter-workspace switch <事项简称> or say 实务级." Load the active matter's matter.md to get matter-specific context and override rules. Write outputs to the matter folder ~/.claude/plugins/config/claude-for-legal-zh/product-legal/matters/<事项简称>/. Never read files from other matters unless 跨事项上下文 is 开.
Purpose
Marketing wants to say the product is the best. Legal needs it to say the truth, or at least not a provable lie. This skill identifies claims that will draw a competitor's cease-and-desist or a regulator's letter of inquiry, and suggests ways to fix the exposure while preserving the force.
Load standards
Read ~/.claude/plugins/config/claude-for-legal-zh/product-legal/CLAUDE.md → ## 营销宣传：
- Comparative advertising policy (permitted with substantiation / discouraged / never)
- Substantiation standards (what is needed before a claim goes out)
- Commonly rejected claims (learn from history)
Retrieve applicable standards before approving copy
Retrieve current advertising and substantiation standards for the applicable jurisdiction and medium (e.g., the Advertising Law, the Anti-Unfair Competition Law, local consumer protection regimes, industry regulator rules, and platform-specific policies). Identify what substantiation thespecific claimneeds—measured by whom, when, sample size, comparable basis—not just whether there issome kind ofsubstantiation on file. Flag implied claims and comparative claims as requiring heightened scrutiny. Check currency: advertising enforcement guidelines and review guidance are updated and continue to evolve. Cite primary sources with precise pinpoints. If you cannot confirm the current standard, flag it for attorney verification—do not state a rule you have not confirmed.
Cite only standards applicable to the specific claim under review. A generic list covering every SAMR guideline, industry rule, or enforcement case makes the load-bearing standards invisible. Do not cite endorsement/testimonial rules unless the copy contains an endorsement, testimonial, or influencer content. Do not cite disclosure overlay rules unless a claim in the material triggers a disclosure overlay. Do not cite an industry regulator unless the copy targets or involves an industry. A standard earns its place in the output because it maps to a specific cited claim; otherwise cut it.
No silent supplementation. If a retrieval query against a configured legal research tool returns few or no results (SAMR rules, advertising violation cases, local consumer protection regulations, industry rules, platform policies), report what was retrieved and stop. Do not fill in from web search or model knowledge without asking. Say: "[tool] search returned [N] results. Coverage on [standard/jurisdiction] appears limited. Options: (1) broaden the search query, (2) try a different research tool, (3) search the web—results will be marked
[联网检索 — 需复核], and should be verified against the issuing authority before relying on them, or (4) mark as unverified and stop. Which do you choose?" It is for the lawyer to decide whether to accept lower-confidence sources.Source attribution tiers. Mark each citation with its source. For model-knowledge citations, use three tiers rather than a single blanket "needs verification" label:
[已确认]— stable, well-known statutory and regulatory citations unlikely to have changed (e.g., Article 9 prohibited words of the Advertising Law, Article 28 false advertising, Article 8 of the Anti-Unfair Competition Law as a concept). These still need verification before copy is approved, but are lower priority.[需验证]— model-knowledge citations that are genuine but should be verified: specific SAMR enforcement actions, advertising violation cases, local consumer protection regulations, industry-specific rules, platform policies, case judgments, thresholds, effective dates, recent updates (advertising enforcement guidelines and endorsement/testimonial disclosure rules are updated frequently).[需精准核实]— precise citations (specific articles and items, judicial interpretation numbers, case docket numbers) carry the highest fabrication risk and should always be verified against the original source.Tool-retrieved citations retain their source label (
[北大法宝]、[威科先行]、[市场监管总局网站]、[平台政策]or MCP tool name); web search citations retain[联网检索 — 需复核]; user-provided citations (from substantiation documents) retain[用户提供]. Tiering makes genuine verification work stand out—someone who verifies everything verifies nothing. Never strip or collapse labels.
Advertising taxonomy
The following categories are structural patterns a reviewer should be able to recognize. Whether a specific wording is actionable depends on the current rules of the applicable jurisdiction, the specific substantiation available, and the audience—search and verify before reaching a conclusion.
Vague/subjective advertising
Subjective assertions with no measurable content. Whether actionable depends on jurisdiction, context, and audience—search before reaching a conclusion.
| Examples |
|---|
| "The best way to manage projects" |
| "You'll love it" |
| "Revolutionary" |
Specific factual advertising
Measurable, specific, and something a reasonable person might rely on.
| Examples | Substantiation to look for |
|---|---|
| "50% faster than [competitor]" | Benchmark data, disclosed methodology, date |
| "Trusted by 10,000 businesses" | Actual number (not cumulative registrations—currenttrust) |
| "Save 5 hours per week" | Research or customer data, disclosed sample |
| "Enterprise-grade security" | What does this mean? What level of protection? Say it clearly, otherwise it is a promise |
| "Personal information protection compliance" | Completed personal information protection impact assessment, actual configuration—this is a contractual commitment |
Comparative advertising (heightened scrutiny)
Name a competitor or allude to a competitor. Before passing, search the applicable rules on comparative advertising in the applicable jurisdiction and medium.
| Example | Revision pattern |
|---|---|
| "Faster than DingTalk" | Either name DingTalk and attach defensible head-to-head data, or abstract it to "faster than traditional chat tools" and attach substantiation |
| "The only platform that can do X" | If anyone else can also do X, it is false—"the first platform to do X..." (if true) or delete "only" |
| "[Competitor] can't do this" | Show your features. Let the audience compare. |
According to ~/.claude/plugins/config/claude-for-legal-zh/product-legal/CLAUDE.md—if comparative claims are "never," flag all. If "allowed with substantiation," check the substantiation.
Implied claims
Not explicitly stated but a reasonable reader would infer it. Search how the applicable advertising regime handles implied claims—implied claims usually carry the same substantiation burden as express claims.
| Example | Implied | Revision |
|---|---|---|
| "Finally, a secure alternative" | Competitors are not secure | "Finally, securely verifiable" |
| Customer logos without context | These companies endorse us | "Customers include..." is fine; "trusted by..." implies more |
| "Built for the healthcare industry" | Complies with personal information protection requirements | Clarify or qualify |
| "Bank-grade security" | Recognized by financial regulators | Either obtain certification or delete |
Absolute claims
No error is allowed. A single counterexample can make them false. Search whether qualifying language can cure the problem under the applicable jurisdiction.
| Example | Revision pattern |
|---|---|
| "Never goes down" | "99.9% uptime" (with an SLA defining that metric) |
| "100% accurate" | A specific, substantiated percentage tied to a benchmark |
| "Guaranteed" | Only if you actually provide a guarantee with terms—this creates warranty liability |
| "Always"/"every" | "Usually"/"most" |
Review
Step 1: Extract every claim
Read the copy. List every sentence or phrase that states a fact, makes a comparison, or makes a promise. Ignore pure puffery in the list.
Step 2: Classify and check
For each claim:
**宣传：**"[精确引用]"
**类型：**[具体事实性 | 比较性 | 暗示性 | 绝对性]
**文件上的证实：**[是 — 链接 | 否 | 未知]
**判断：**[✅ 没问题 | ⚠️ 需要证实 | ⚠️ 需要改写 | 🔴 删除]
**建议修改：**"[在保持力度的同时替代措辞]"
**理由：**[一句话]
Step 3: Test against the product
Can the product really do what the copy says? This is not a philosophical question—check the PRD or ask the product manager.
Common deviation: marketing copy was written based on early specifications, the product has changed, and no one updated the copy.
Step 4: Output
Prefix with the work product header in ~/.claude/plugins/config/claude-for-legal-zh/product-legal/CLAUDE.md ## 输出规范 (varies by user role—see ## 使用者）。
[工作成果页眉 — 按插件配置 ## 输出规范]
# 营销审查：[活动/素材名称]
**审查日期：**[日期]
**素材：**[落地页/邮件/广告/等]
---
## 摘要
[N]条宣传已审查。[N]✅ [N]⚠️ [N]🔴
**可发布：**[是 | 经以下修改后 | 否——需重写]
> **在对"可发布：是"（即批准某宣传用于外部发布）输出前：** 读取 `~/.claude/plugins/config/claude-for-legal-zh/product-legal/CLAUDE.md` 中的 `## 使用者`。如果角色为非法务人员：
>
> > 批准一项营销宣传对外发布是一项法律行为——一旦发布，证实缺口和比较性宣传曝光即成为执法或竞争对手挑战风险。您是否已与律师审查？如已审查，继续。如未审查，以下是带给律师的简要说明：
> >
> > [生成1页摘要：素材、已批准宣传、宣传类型（具体事实性/比较性/暗示性/绝对性）、每条的文件证实、任何标记的暗示性宣传，以及文案上线前向律师提出的三个问题。]
> >
> > 如需要寻找律师：中华全国律师协会或所在地地方律师协会的推荐服务是最快的起点。
>
> 在未获得明确同意前，不越过此准入口输出"可发布：是"。"经以下修改后"和"否——需重写"不需要准入——这些是审查判断，不是批准。
---
## 逐条分析
[所有第2步的宣传块，按以下分组：🔴 在最前，然后是 ⚠️，然后是 ✅]
---
## 建议修改稿
[对于短素材——50字以内，或一条推文、标题、一句话、标语、短广告——本节输出为实际修改后的文案，修改已内联应用，而非描述改了什么。读者应能复制粘贴本节到素材中。
对于较长素材（>50字但<300字），展示修改后的文案，修改已内联应用。
对于更长素材（300字以上），以分项差异摘要方式总结修改（"删除宣传1。将宣传3改写为删除'任何'。为受监管领域风险将宣传4弱化。"），而非粘贴整个素材。
对短素材，修改的元描述永远不是可接受的输出——当素材是一行时，输出应就是修改后的一行。]
---
## 发布前需要的证实
| 宣传 | 需要 | 从谁 |
|---|---|---|
| [宣传] | [数据类型] | [产品经理/数据团队/开发] |
---
## 引用检查
本审查中引用的任何市场监管总局规则、广告违法案例、地方消费者保护法规、行业法规或平台政策均由AI模型生成且未经原始来源验证。在依赖特定规则来通过或拒绝文案之前，对照法律研究工具（北大法宝、威科先行、法信或您的律所研究平台）核实其准确性和当前生效日期——广告执法指南、平台规则和地方消费者保护制度均频繁更新。每条引用上的来源标签（如 `[市场监管总局网站]`、`[联网检索 — 需复核]`）显示其来源；`需验证` 标签具有较高的编造风险，应首先检查。
Disclosure coverage
Copy that involves any of the following fact patterns is subject to additional disclosure regimes. Search the current disclosure requirements of the applicable jurisdiction (including any platform policies and industry-specific rules) and verify timeliness—these regimes update frequently.
- Testimonials/reviews— Material connections between speakers and advertisers generally require disclosure; look up current form and placement rules
- Influencer content— Look up current labeling, clarity, and prominence requirements for this channel and audience
- "Results vary"/atypical results— When displayed results are not representative, look up whether disclosure is required (and in what form)
- Free trials/auto-renewal/negative option— Look up current prominence and consent requirements for auto-conversion terms. Under Chinese law, pay attention to the requirements of the Consumer Rights Protection Law and the E-Commerce Law on auto-renewal
Close with the next-step decision tree
Close with the next-step decision tree in CLAUDE.md ## 输出规范 . Customize the options to what this skill has just produced—the five default branches (draft X, escalate, supplement facts, monitor and wait, other) are a starting point, not locked in. The decision tree is the output; the lawyer makes the choice.
What this skill does not do
- It does not write marketing copy. It fixes the parts of the copy that are wrong. Suggested revisions stay forceful, but marketers own the feel for language.
- It does not substantiate claims. It identifies which ones need substantiation and who has the data.
- It does not review design or images—only text. If an image implies a claim (a red X over a competitor's logo), flag it, but visual review is a human judgment.