Can I use both CLAUDE.md and AGENTS.md at the project root?

No, the core principle is to maintain only one root context file. You should choose either CLAUDE.md or AGENTS.md at the root, while using child AGENTS.md files for subdirectories.

Does this skill help with existing projects?

Yes, it includes a detection workflow to analyze existing repositories, identify gaps in documentation infrastructure, and suggest maintenance or audit actions.

What is an Intent Layer?

An Intent Layer is a hierarchical system of documentation files (AGENTS.md) that provides AI agents with high-level architectural context, invariants, and navigation rules for a codebase.

How does the skill prevent AI context window issues?

The skill includes scripts to measure directory complexity and token counts, ensuring individual documentation nodes remain under 4k tokens for efficient processing.

When should I create a child AGENTS.md node?

Child nodes are recommended when a directory exceeds 20k tokens, when there is a significant shift in responsibility, or when hidden invariants need local documentation.

Intent Layer Architect

Name: Intent Layer Architect
Author: Bobxsszef

byBobxsszef

•

Learning & Documentation

Deploys hierarchical AGENTS.md infrastructure to provide AI agents with high-level architectural context and senior-level navigation paths.

The Intent Layer skill transforms complex codebases into AI-friendly environments by implementing a structured, hierarchical documentation system using AGENTS.md files. It automates the process of analyzing project structure, measuring token density, and establishing clear boundaries for AI agents, ensuring they understand invariants, ownership, and cross-cutting concerns just like a senior engineer. This skill prevents context overflow by strategically placing child nodes only where needed, streamlining AI-assisted development through standardized templates and specialized capture protocols for subject matter expertise.

Key Features

011 GitHub stars

02Hierarchical AGENTS.md generation for scalable context management

03Token-density measurement to prevent AI context window overflow

04Automated state detection and structural analysis of complex codebases

05Maintenance mode for auditing nodes and capturing SME knowledge

06Standardized templates for root and subdirectory documentation nodes

Use Cases

01Onboarding AI agents to large, complex legacy repositories

02Restructuring existing CLAUDE.md files into a multi-layered intent system

03Initializing architectural documentation for a new software project

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add bobxsszef/skills intent-layer

For use in Claude.ai and ChatGPT

Download Skill