business-brain

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Business Brain Pattern

Business Brain模式

This skill implements the "Business Brain Pattern" for AI coding assistants. It centralizes all shared business and brand knowledge in one location—the "brain"—and allows individual skills or agent queries to reference it selectively rather than duplicating it.

本技能为AI编码助手实现了“Business Brain模式”。它将所有共享的业务和品牌知识集中存储在一个位置——即“大脑”——并允许各个技能或Agent查询选择性地引用这些知识，而非重复存储。

The Problem

问题

Loading full brand guides, audience profiles, and positioning documents into every skill or prompt bloats the context window. This increases token usage, diffuses the agent's attention from the actual task instructions, and creates a maintenance nightmare when brand guidelines change.

将完整的品牌指南、受众资料和定位文档加载到每个技能或提示词中会导致上下文窗口臃肿。这会增加token使用量，分散Agent对实际任务指令的注意力，并且当品牌准则变更时会带来维护难题。

The Solution

解决方案

The Business Brain pattern keeps individual skill prompts lean. Each skill loads only the brain sections it actually uses. The brain is organized into modular files by topic.

Business Brain模式让单个技能的提示词保持精简。每个技能仅加载其实际需要的“大脑”部分内容。“大脑”按主题组织为模块化文件。

Usage

使用方法

When tasked with retrieving brand context, writing on-brand copy, or answering questions about the project's audience:

Check for a local brain: Look for an
```
.agent/brain/
```
directory (or legacy
```
brain/
```
,
```
.claude/brain/
```
,
```
.gemini/brain/
```
) in the project workspace.
Selective loading: Use your file reading tools to load only the specific files needed for the task.
- Example: A social media drafting task might only need
```
brand-voice.md
```
  and
```
audience-profiles.md
```
  .
- Example: A technical spec task might only need
```
content-rules.md
```
  .
DO NOT load the entire brain directory if only one piece of context is needed. Trust the modular structure.

当需要获取品牌背景、撰写符合品牌风格的文案，或回答关于项目受众的问题时：

检查本地大脑：在项目工作区中查找
```
.agent/brain/
```
目录（或旧版的
```
brain/
```
、
```
.claude/brain/
```
、
```
.gemini/brain/
```
）。
选择性加载：使用文件读取工具仅加载任务所需的特定文件。
- 示例：社交媒体文案撰写任务可能只需要
```
brand-voice.md
```
  和
```
audience-profiles.md
```
  。
- 示例：技术规格文档任务可能只需要
```
content-rules.md
```
  。
请勿在仅需单一上下文内容时加载整个大脑目录。请信任模块化结构。

Initialization

初始化

If the user wants to set up a new Business Brain for their project, run the included scaffolding script to generate the template files:

bash

bash ~/.agents/skills/business-brain/scripts/init-brain.sh

(Path may vary depending on the agent environment, e.g.,
~/.claude/skills/
or
~/.gemini/skills/
)

This script creates an

.agent/brain/

directory with the following templates:

```
brand-voice.md
```
```
audience-profiles.md
```
```
positioning.md
```
```
content-rules.md
```

如果用户想要为其项目搭建新的Business Brain，运行附带的脚手架脚本以生成模板文件：

bash

bash ~/.agents/skills/business-brain/scripts/init-brain.sh

(路径可能因Agent环境而异，例如
~/.claude/skills/
或
~/.gemini/skills/
)

该脚本会创建一个

.agent/brain/

目录，并包含以下模板文件：

```
brand-voice.md
```
```
audience-profiles.md
```
```
positioning.md
```
```
content-rules.md
```

Structure of the Brain

大脑的结构

A well-built business brain isn't a data dump. It's structured and scannable:

一个完善的Business Brain并非数据堆砌，而是结构化且易于浏览的：

brand-voice.md

brand-voice.md

Specify sentence length preferences.
Define the register (first person, formal "we", conversational "I").
List words/phrases to always use and never use.
Describe the energy level.
Include 2-3 sample paragraphs demonstrating the voice, and examples of off-brand writing.

指定句子长度偏好。
定义语体（第一人称、正式的“我们”、口语化的“我”）。
列出始终使用和绝对禁用的词汇/短语。
描述语气活力水平。
包含2-3段展示品牌语气的示例段落，以及不符合品牌风格的写作示例。

audience-profiles.md

audience-profiles.md

Define 2-4 primary audience segments.
Job title/life situation.
What they care about most and what they're skeptical of.
Familiarity level with the topic.
Language/framing that resonates.

定义2-4个主要受众群体。
职位头衔/生活场景。
他们最关心的内容以及持怀疑态度的方面。
对主题的熟悉程度。
能引起共鸣的语言/表述方式。

positioning.md

positioning.md

What the company/product does.
Who it's for.
Why someone would choose it over alternatives.
Explicitly state what you do not claim. Keep it focused (200-300 words).

公司/产品的业务范围。
目标用户群体。
用户选择该产品而非竞品的原因。
明确说明不做哪些宣称。保持内容聚焦（200-300字）。

content-rules.md

content-rules.md

Structural preferences across output types.
Header formatting conventions.
List vs. prose preferences.
Length guidelines.
How to handle calls to action.

不同输出类型的结构偏好。
标题格式规范。
列表与散文的偏好选择。
篇幅指南。
行动号召的处理方式。

Agentic OS Integration

Agentic OS集成

If the current project root contains an

.agent/

directory, this skill MUST participate in the Agentic OS shared-memory model.

At the end of your execution, check for

.agent/state/last-run.json

. If it exists, append or update the file using its required schema to log your run. Ensure you capture your runtime (

agent_id

assigned_skill

, a concise

description

decision_log

, and

outcome

. Use

python3 ~/.agents/skills/heartbeat/scripts/heartbeat.py complete <task_id> ...

if completing a task from the queue.

如果当前项目根目录包含

.agent/

目录，本技能必须参与Agentic OS共享内存模型。

在执行结束时，检查

.agent/state/last-run.json

文件。若该文件存在，请按照其要求的架构追加或更新文件，以记录本次运行信息。确保捕获运行时信息（

agent_id

）、

assigned_skill

、简洁的

description

、

decision_log

和

outcome

。若完成队列中的任务，请使用

python3 ~/.agents/skills/heartbeat/scripts/heartbeat.py complete <task_id> ...

命令。