intent-layer 意图层：分层式 AI 智能体代码库导航基础设施 | AGENTS.md 最佳实践

intent-layer by crafter-station/skills

190 周安装量

39 GitHub Stars

GitHub

安装命令

npx skills add https://github.com/crafter-station/skills --skill intent-layer

AI/机器学习软件工程代码质量

🇨🇳中文介绍

意图层

分层式 AGENTS.md 基础设施，让智能体像资深工程师一样浏览代码库。

核心原则

只有一个根上下文文件。 项目根目录下不应同时存在 CLAUDE.md 和 AGENTS.md。鼓励在子目录中创建子级 AGENTS.md 以处理复杂的子系统。

工作流程

1. 检测状态
   scripts/detect_state.sh /path/to/project
   → 返回：none | partial | complete

2. 路由
   none/partial → 初始设置（步骤 3-5）
   complete     → 维护（步骤 6）

3. 测量 [关卡 - 先显示表格]
   scripts/analyze_structure.sh /path/to/project
   scripts/estimate_tokens.sh /path/to/each/source/dir

4. 决策
   无根文件  → 询问：CLAUDE.md 还是 AGENTS.md？
   有根文件  → 添加意图层部分 + 必要时添加子节点

5. 执行
   使用 references/templates.md 获取结构
   使用 references/node-examples.md 获取真实世界模式
   验证：一个根节点，READ-FIRST 指令，每个节点 <4k tokens

6. 维护模式（当状态=complete时）
   询问用户：
   a) 审计节点     → 使用 references/capture-protocol.md 进行领域专家提问
   b) 寻找候选节点 → 重新测量 tokens，建议新节点
   c) 两者都做

何时创建子节点

信号	操作
目录中 >20k tokens	创建 AGENTS.md
职责转移	创建 AGENTS.md
隐藏的契约/不变量

广告位招租

在这里展示您的产品或服务

触达数万 AI 开发者，精准高效

联系我们

🇺🇸English

Intent Layer

Hierarchical AGENTS.md infrastructure so agents navigate codebases like senior engineers.

Core Principle

Only ONE root context file. CLAUDE.md and AGENTS.md should NOT coexist at project root. Child AGENTS.md in subdirectories are encouraged for complex subsystems.

Workflow

1. Detect state
   scripts/detect_state.sh /path/to/project
   → Returns: none | partial | complete

2. Route
   none/partial → Initial setup (steps 3-5)
   complete     → Maintenance (step 6)

3. Measure [gate - show table first]
   scripts/analyze_structure.sh /path/to/project
   scripts/estimate_tokens.sh /path/to/each/source/dir

4. Decide
   No root file  → Ask: CLAUDE.md or AGENTS.md?
   Has root file → Add Intent Layer section + child nodes if needed

5. Execute
   Use references/templates.md for structure
   Use references/node-examples.md for real-world patterns
   Validate: one root, READ-FIRST directive, <4k tokens per node

6. Maintenance mode (when state=complete)
   Ask user:
   a) Audit nodes     → Use references/capture-protocol.md for SME questions
   b) Find candidates → Re-measure tokens, suggest new nodes
   c) Both

When to Create Child Nodes

Signal	Action

20k tokens in directory | Create AGENTS.md
Responsibility shift | Create AGENTS.md
Hidden contracts/invariants | Document in nearest ancestor
Cross-cutting concern | Place at LCA

Do NOT create for: every directory, simple utilities, test folders (unless complex).

Capture Questions

When documenting existing code, ask:

What does this area own? What's out of scope?
What invariants must never be violated?
What repeatedly confuses new engineers?
What patterns should always be followed?

Resources

Scripts:

scripts/detect_state.sh - Check Intent Layer state (none/partial/complete)
scripts/analyze_structure.sh - Find semantic boundaries
scripts/estimate_tokens.sh - Measure directory complexity

References:

references/templates.md - Root and child node templates
references/node-examples.md - Real-world examples
references/capture-protocol.md - SME interview protocol

Weekly Installs

115

Repository

crafter-station/skills

GitHub Stars

First Seen

Jan 19, 2026

Security Audits

Gen Agent Trust HubPass SocketPass SnykPass

Installed on

claude-code91

gemini-cli87

codex83

opencode83

cursor70

github-copilot69