brand-voice-enforcement

Anthropic ｜ 行銷與 SEO ｜ 原文 ｜ Apache-2.0

把品牌指南的語氣與用語套到銷售行銷內容的產出上

---
name: brand-voice-enforcement
description: >
  這個 skill 把品牌指南套用到內容創作上。當使用者要求「寫一封 email」「草擬提案」「做 pitch deck」
  「寫 LinkedIn 貼文」「草擬簡報」「寫 Slack 訊息」「草擬銷售內容」，或任何需要套用品牌語氣的內容創作請求時使用。
  也會在出現「on-brand」「brand voice」「enforce voice」「apply brand guidelines」「brand-aligned content」
  「write in our voice」「use our brand tone」「make this sound like us」「rewrite this in our tone」
  或「this doesn't sound on-brand」時觸發。不適用於從零產生指南（用 guideline-generation）
  或探索品牌素材（用 discover-brand）。
---

# Brand Voice Enforcement

把現有的品牌指南套用到所有銷售與行銷內容的產生上。載入使用者的品牌指南，把 voice constants 和 tone flexes 套到內容請求上，驗證輸出，並說明品牌上的選擇。

## 載入品牌指南

用下面的順序找使用者的品牌指南，找到就停：

1. **Session context** — 檢查這個 session 稍早是否已經產生過品牌指南（透過 `/brand-voice:generate-guidelines`）。如果有，它們已經在對話裡，直接用。session 裡產生的指南最新，也最能反映使用者最近的意圖。

2. **本地指南檔** — 檢查使用者工作資料夾裡有沒有 `.claude/brand-voice-guidelines.md`。不要用相對於 agent 目前工作目錄的相對路徑，因為在 Cowork 裡 agent 是從 plugin cache 目錄執行的，不是使用者的專案。要相對於使用者的工作資料夾解析路徑。如果沒有設定工作資料夾，就跳過這一步。

3. **問使用者** — 如果上面都沒找到指南，告訴使用者：
   「我找不到你的品牌指南。你可以：
   - 執行 `/brand-voice:discover-brand`，在你的各個平台上找品牌素材
   - 執行 `/brand-voice:generate-guidelines`，從文件或逐字稿建立指南
   - 直接把指南貼到這個對話，或告訴我檔案在哪」

   等使用者提供指南再繼續。

另外也要讀 `.claude/brand-voice.local.md` 取得執行強度的設定（就算指南是從別的地方來的）：
- `strictness`: strict | balanced | flexible
- `always-explain`: 是否永遠說明品牌上的選擇

## 執行流程

### 1. 分析內容請求

動筆之前先確認：
- **內容類型**：email、簡報、提案、社群貼文、訊息等
- **目標讀者**：角色、資歷、產業、公司階段
- **需要的關鍵訊息**：適用哪些 message pillar
- **特定要求**：長度、格式、語氣覆寫

### 2. 套用 Voice Constants

Voice 是品牌的個性，在所有內容裡都維持不變：
- 套用指南裡的「We Are / We Are Not」屬性
- 一致地使用品牌個性
- 採用核准的用語，拒絕禁用的詞
- 遵循訊息框架和價值主張

「voice constant, tone flexes」模型請參考 `references/voice-constant-tone-flexes.md`。

### 3. 依情境調整 Tone

Tone 會依內容類型和讀者調整。用指南裡的 tone-by-context 矩陣來設定：
- **正式度**：這份該多正式或多隨性？
- **能量**：要有多少急迫感或熱情？
- **技術深度**：要多詳細，或多好懂？

### 4. 產生內容

產出的內容要：
- 全程符合品牌 voice 屬性
- 遵循這個內容類型專屬的 tone 指引
- 自然地帶入關鍵訊息（不要硬塞）
- 使用偏好的用語
- 品質與風格要貼近指南裡的範例

複雜或長篇的內容，交給 content-generation agent（定義在 `agents/content-generation.md`）。
重要性高的內容，交給 quality-assurance agent（定義在 `agents/quality-assurance.md`）做驗證。

### 5. 驗證並說明

內容產生後：
- 簡短點出套用了哪些品牌指南
- 說明關鍵的 voice 與 tone 決定
- 註明有哪些地方為了情境而調整了指南
- 主動提議根據回饋再修

設定裡的 `always-explain` 為 true 時，每次回覆都要附上品牌套用說明。

## 處理衝突

使用者的請求跟品牌指南衝突時：
1. 清楚說明衝突在哪
2. 給出建議
3. 提供選項：嚴格遵循指南、依情境調整，或直接覆寫

預設是調整指南，並說明取捨。

## Open Questions 意識

Open questions 是產生指南時標記出來、還沒解決的品牌定位決策，存在指南的「Open Questions」段落裡。產生內容時，要檢查品牌指南有沒有 open questions：
- 如果內容碰到還沒解決的 open question，要註明
- 除非使用者另有指示，否則採用 agent 在該 open question 上的建議
- 如果該問題對內容影響很大，建議使用者把它解決掉

## 參考檔案

- **`references/voice-constant-tone-flexes.md`** — 「voice constant, tone flexes」心智模型、「We Are / We Are Not」表格結構，以及 tone-by-context 矩陣的說明
- **`references/before-after-examples.md`** — 各內容類型的前後對照範例，展示實際執行的樣子

---
name: brand-voice-enforcement
description: >
  This skill applies brand guidelines to content creation. It should be used when
  the user asks to "write an email", "draft a proposal", "create a pitch deck",
  "write a LinkedIn post", "draft a presentation", "write a Slack message",
  "draft sales content", or any content creation request where brand voice should
  be applied. Also triggers on "on-brand", "brand voice", "enforce voice",
  "apply brand guidelines", "brand-aligned content", "write in our voice",
  "use our brand tone", "make this sound like us", "rewrite this in our tone",
  or "this doesn't sound on-brand". Not for generating guidelines from scratch
  (use guideline-generation) or discovering brand materials (use discover-brand).
---

# Brand Voice Enforcement

Apply existing brand guidelines to all sales and marketing content generation. Load the user's brand guidelines, apply voice constants and tone flexes to the content request, validate output, and explain brand choices.

## Loading Brand Guidelines

Find the user's brand guidelines using this sequence. Stop as soon as you find them:

1. **Session context** — Check if brand guidelines were generated earlier in this session (via `/brand-voice:generate-guidelines`). If so, they are already in the conversation. Use them directly. Session-generated guidelines are the freshest and reflect the user's most recent intent.

2. **Local guidelines file** — Check for `.claude/brand-voice-guidelines.md` inside the user's working folder. Do NOT use a relative path from the agent's current working directory — in Cowork, the agent runs from a plugin cache directory, not the user's project. Resolve the path relative to the user's working folder. If no working folder is set, skip this step.

3. **Ask the user** — If none of the above found guidelines, tell the user:
   "I couldn't find your brand guidelines. You can:
   - Run `/brand-voice:discover-brand` to find brand materials across your platforms
   - Run `/brand-voice:generate-guidelines` to create guidelines from documents or transcripts
   - Paste guidelines directly into this chat or point me to a file"

   Wait for the user to provide guidelines before proceeding.

Also read `.claude/brand-voice.local.md` for enforcement settings (even if guidelines came from another source):
- `strictness`: strict | balanced | flexible
- `always-explain`: whether to always explain brand choices

## Enforcement Workflow

### 1. Analyze the Content Request

Before writing, identify:
- **Content type**: email, presentation, proposal, social post, message, etc.
- **Target audience**: role, seniority, industry, company stage
- **Key messages needed**: which message pillars apply
- **Specific requirements**: length, format, tone overrides

### 2. Apply Voice Constants

Voice is the brand's personality — it stays constant across all content:
- Apply "We Are / We Are Not" attributes from guidelines
- Use brand personality consistently
- Incorporate approved terminology; reject prohibited terms
- Follow messaging framework and value propositions

Refer to `references/voice-constant-tone-flexes.md` for the "voice constant, tone flexes" model.

### 3. Flex Tone for Context

Tone adapts by content type and audience. Use the tone-by-context matrix from guidelines to set:
- **Formality**: How formal or casual should this be?
- **Energy**: How much urgency or enthusiasm?
- **Technical depth**: How detailed or accessible?

### 4. Generate Content

Create content that:
- Matches brand voice attributes throughout
- Follows tone guidelines for this specific content type
- Incorporates key messages naturally (not forced)
- Uses preferred terminology
- Mirrors the quality and style of guideline examples

For complex or long-form content, delegate to the content-generation agent (defined in `agents/content-generation.md`).
For high-stakes content, delegate to the quality-assurance agent (defined in `agents/quality-assurance.md`) for validation.

### 5. Validate and Explain

After generating content:
- Briefly highlight which brand guidelines were applied
- Explain key voice and tone decisions
- Note any areas where guidelines were adapted for context
- Offer to refine based on feedback

When `always-explain` is true in settings, include brand application notes with every response.

## Handling Conflicts

When the user's request conflicts with brand guidelines:
1. Explain the conflict clearly
2. Provide a recommendation
3. Offer options: follow guidelines strictly, adapt for context, or override

Default to adapting guidelines with an explanation of the tradeoff.

## Open Questions Awareness

Open questions are unresolved brand positioning decisions flagged during guideline generation, stored in the guidelines under an "Open Questions" section. When generating content, check if the brand guidelines contain open questions:
- If content touches an unresolved open question, note it
- Apply the agent's recommendation from the open question unless the user specifies otherwise
- Suggest resolving the question if it significantly impacts the content

## Reference Files

- **`references/voice-constant-tone-flexes.md`** — The "voice constant, tone flexes" mental model, "We Are / We Are Not" table structure, and tone-by-context matrix explanation
- **`references/before-after-examples.md`** — Before/after content examples per content type showing enforcement in practice