Context Hierarchy Design Skill

上下文层级设计技能

Design a memory hierarchy that loads context progressively based on task needs.

设计一个根据任务需求渐进式加载上下文的内存层级结构。

Purpose

用途

Not all context is needed all the time. A well-designed hierarchy ensures agents get exactly the context they need without bloat.

并非所有上下文始终都是必需的。一个设计良好的层级结构可确保Agent获得其所需的准确上下文，而不会产生冗余。

When to Use

适用场景

Setting up a new project's context infrastructure
Refactoring bloated CLAUDE.md files
Creating task-specific context loading
Optimizing agent startup time
Scaling context for multiple task types

搭建新项目的上下文基础设施
重构臃肿的CLAUDE.md文件
创建特定任务的上下文加载机制
优化Agent启动时间
为多种任务类型扩展上下文

The Three-Tier Memory Strategy

三级内存策略

text

Tier 1: CLAUDE.md (minimal, always loaded)
        |
        v
Tier 2: Priming Commands (task-specific, on-demand)
        |
        v
Tier 3: File Reads (just-in-time, as needed)

text

Tier 1: CLAUDE.md (minimal, always loaded)
        |
        v
Tier 2: Priming Commands (task-specific, on-demand)
        |
        v
Tier 3: File Reads (just-in-time, as needed)

Design Process

设计流程

Step 1: Audit Current State

步骤1：审计当前状态

Analyze existing context infrastructure:

Measure CLAUDE.md size
Count imports and their sizes
Review command coverage
Identify task types

text

Checklist:
- [ ] CLAUDE.md token count
- [ ] Import count and total
- [ ] Task types identified
- [ ] Command coverage mapped

分析现有上下文基础设施：

测量CLAUDE.md的大小
统计导入文件及其大小
检查指令覆盖范围
识别任务类型

text

Checklist:
- [ ] CLAUDE.md token count
- [ ] Import count and total
- [ ] Task types identified
- [ ] Command coverage mapped

Step 2: Categorize Context

步骤2：分类上下文

For each piece of existing context, ask:

Question	If Yes ->
Needed for EVERY task?	Tier 1 (CLAUDE.md)
Needed for task TYPE?	Tier 2 (Priming)
Needed for specific work?	Tier 3 (On-demand)

对于每一段现有上下文，思考：

问题	如果是 ->
所有任务都需要吗？	Tier 1（CLAUDE.md）
特定任务类型需要吗？	Tier 2（Priming）
特定工作场景需要吗？	Tier 3（按需加载）

Step 3: Design Tier 1 (CLAUDE.md)

步骤3：设计Tier 1（CLAUDE.md）

Minimal essentials only:

markdown

undefined

仅保留最基础的必要内容：

markdown

undefined

Project Name

Context

One-sentence project description.

Tooling

Language: X
Runtime: Y
Package Manager: Z

Language: X
Runtime: Y
Package Manager: Z

Key Commands

```
cmd test
```
- Run tests
```
cmd build
```
- Build project

```
cmd test
```
- Run tests
```
cmd build
```
- Build project

Critical Rules

Most important rule
Second most important
Third most important (max 5)


Target size: less than 2KB.

Most important rule
Second most important
Third most important (max 5)


目标大小：小于2KB。

Step 4: Design Tier 2 (Priming Commands)

步骤4：设计Tier 2（Priming Commands）

Create commands for each task type:

Task Type	Command	Context Loaded
General	/prime	README, git status
Bug Fix	/prime-bug	Recent commits, test files
Feature	/prime-feature	Architecture, API patterns
Review	/prime-review	Style guide, PR diff
Docs	/prime-docs	Existing docs, API surface

Priming Command Template:

markdown

undefined

为每种任务类型创建指令：

任务类型	指令	加载的上下文
通用	/prime	README、git状态
修复Bug	/prime-bug	最近提交记录、测试文件
开发功能	/prime-feature	架构、API模式
代码评审	/prime-review	风格指南、PR差异
文档编写	/prime-docs	现有文档、API接口

Priming指令模板：

markdown

undefined

Prime: [Task Type]

Run

[Discovery commands for current state]

Read

[Essential files for this task type]

Report

[Summary prompt for understanding]

undefined

[Summary prompt for understanding]

undefined

Step 5: Design Tier 3 (On-Demand)

步骤5：设计Tier 3（按需加载）

Identify files loaded during work:

Category	Loading Strategy
Source files	Read specific file when editing
Test files	Read when running tests
Configs	Read when configuring
Dependencies	Read when debugging deps

Principle: Never pre-load what can be loaded on-demand.

识别工作过程中需要加载的文件：

类别	加载策略
源文件	编辑时读取特定文件
测试文件	运行测试时读取
配置文件	配置时读取
依赖文件	调试依赖时读取

原则： 绝不预加载可按需加载的内容。

Output Format

输出格式

Design document structure:

markdown

undefined

设计文档结构：

markdown

undefined

Context Hierarchy Design: [Project]

Tier 1: CLAUDE.md (Always Loaded)

Target Size: <2KB Content:

Project context (1 sentence)
Tooling (language, runtime, PM)
Key commands (3-5)
Critical rules (3-5)

Target Size: <2KB Content:

Project context (1 sentence)
Tooling (language, runtime, PM)
Key commands (3-5)
Critical rules (3-5)

Tier 2: Priming Commands (Task-Specific)

/prime

Purpose: General codebase understanding Loads: README, git status, project structure

/prime-bug

Purpose: Bug fixing context Loads: Recent commits, test patterns, error logs

/prime-feature

Purpose: Feature development context Loads: Architecture, API patterns, similar features

Tier 3: On-Demand (Just-in-Time)

Strategy: Load during execution, not pre-emptively

Source files: Read when editing
Tests: Read when testing
Configs: Read when configuring

Strategy: Load during execution, not pre-emptively

Source files: Read when editing
Tests: Read when testing
Configs: Read when configuring

Migration Plan

Design Patterns

设计模式

Pattern: Import Hierarchy

模式：导入层级结构

text

CLAUDE.md (always)
  -> imports/core.md (always, if large)
  -> imports/tooling.md (conditional)
  -> imports/workflows.md (conditional)

text

CLAUDE.md (always)
  -> imports/core.md (always, if large)
  -> imports/tooling.md (conditional)
  -> imports/workflows.md (conditional)

Pattern: Directory-Scoped Memory

模式：目录范围内存

text

/project
  CLAUDE.md (project-wide)
  /frontend
    CLAUDE.md (frontend-specific)
  /backend
    CLAUDE.md (backend-specific)

text

/project
  CLAUDE.md (project-wide)
  /frontend
    CLAUDE.md (frontend-specific)
  /backend
    CLAUDE.md (backend-specific)

Pattern: Task-Type Priming

模式：任务类型预启动

text

/prime -> Base context
/prime-{type} -> Type-specific context

text

/prime -> Base context
/prime-{type} -> Type-specific context

Validation Criteria

验证标准

After design, verify:

CLAUDE.md under 2KB
Every task type has priming command
No "just in case" loading
Progressive disclosure works
Common workflows smooth

设计完成后，验证以下内容：

CLAUDE.md大小小于2KB
每种任务类型都有对应的Priming指令
不存在“以防万一”的加载
渐进式披露机制有效
常见工作流程顺畅

Anti-Patterns to Avoid

需避免的反模式

Anti-Pattern	Problem	Solution
Everything in CLAUDE.md	Bloat	Move to priming
No priming commands	Static context	Add task-type priming
Import everything	Wasted tokens	Conditional imports
Directory CLAUDE.md bloat	Compounding	Keep scoped minimal

反模式	问题	解决方案
所有内容都放在CLAUDE.md中	冗余臃肿	迁移到Priming指令
无Priming指令	上下文静态不变	添加任务类型专属Priming
导入所有内容	浪费Token	使用条件导入
目录级CLAUDE.md臃肿	冗余叠加	保持范围内容极简

Key Quote

关键引用

"Use context priming over CLAUDE.md or similar auto-loading memory files. Engineering work constantly changes, but memory files only grow."

“优先使用上下文Priming，而非CLAUDE.md或类似的自动加载内存文件。工程工作内容不断变化，但内存文件只会不断膨胀。”

Cross-References

交叉引用

@context-priming-patterns.md - Priming command patterns
@rd-framework.md - Reduce unnecessary context
@context-layers.md - What loads into context
@minimum-context-principle.md - Include only what's necessary

@context-priming-patterns.md - Priming指令模式
@rd-framework.md - 减少不必要的上下文
@context-layers.md - 上下文加载内容
@minimum-context-principle.md - 仅保留必要内容

Version History

版本历史

v1.0.0 (2025-12-26): Initial release

v1.0.0 (2025-12-26): 初始版本

Last Updated

最后更新

Date: 2025-12-26 Model: claude-opus-4-5-20251101

日期: 2025-12-26 模型: claude-opus-4-5-20251101

context-hierarchy-design

Original

Translation

Context Hierarchy Design Skill

上下文层级设计技能

Purpose

用途

When to Use

适用场景

The Three-Tier Memory Strategy

三级内存策略

Design Process

设计流程

Step 1: Audit Current State

步骤1：审计当前状态

Step 2: Categorize Context

步骤2：分类上下文

Step 3: Design Tier 1 (CLAUDE.md)

步骤3：设计Tier 1（CLAUDE.md）

Project Name

Project Name

Context

Context

Tooling

Tooling

Key Commands

Key Commands

Critical Rules

Critical Rules

Step 4: Design Tier 2 (Priming Commands)

步骤4：设计Tier 2（Priming Commands）

Prime: [Task Type]

Prime: [Task Type]

Run

Run

Read

Read

Report

Report

Step 5: Design Tier 3 (On-Demand)

步骤5：设计Tier 3（按需加载）

Output Format

输出格式

Context Hierarchy Design: [Project]

Context Hierarchy Design: [Project]

Tier 1: CLAUDE.md (Always Loaded)

Tier 1: CLAUDE.md (Always Loaded)

Tier 2: Priming Commands (Task-Specific)

Tier 2: Priming Commands (Task-Specific)

/prime

/prime

/prime-bug

/prime-bug

/prime-feature

/prime-feature

Tier 3: On-Demand (Just-in-Time)

Tier 3: On-Demand (Just-in-Time)

Migration Plan

Migration Plan

Design Patterns

设计模式

Pattern: Import Hierarchy

模式：导入层级结构

Pattern: Directory-Scoped Memory

模式：目录范围内存

Pattern: Task-Type Priming

模式：任务类型预启动

Validation Criteria

验证标准

Anti-Patterns to Avoid

需避免的反模式

Key Quote

关键引用

Cross-References

交叉引用

Version History

版本历史

Last Updated

最后更新