How many instructions should a CLAUDE.md file contain?

To maintain high reliability, target 50-80 instructions for project-level files and 30-40 for user-level files, keeping the combined total under 150.

What is the WHAT/WHY/HOW framework?

It is a structured approach for project configs: WHAT defines the purpose/stack, WHY explains architectural decisions, and HOW lists essential commands.

Why shouldn't I use Claude as a linter?

LLMs follow style rules inconsistently. Using deterministic tools like ESLint or Prettier is more reliable and preserves Claude's instruction budget for more complex tasks.

Where are user-level configurations stored?

User-level configurations, which apply to all projects, should be placed in the ~/.claude/CLAUDE.md directory.

What is the benefit of a CLAUDE.md file?

It serves as the primary context transfer point, providing Claude with the essential project purpose, tech stack, and commands it needs to work effectively.

CLAUDE.md Configuration Architect

Name: CLAUDE.md Configuration Architect
Author: dhruvbaldawa

bydhruvbaldawa

•

Productivity & Workflow

Generates and audits high-leverage CLAUDE.md configuration files to optimize context transfer and model performance.

About

This skill empowers developers to master Claude Code's configuration by creating and reviewing CLAUDE.md files based on HumanLayer guidelines. It enforces critical best practices such as the WHAT/WHY/HOW framework, strict instruction budgeting, and progressive disclosure. By identifying anti-patterns—like using Claude as a linter for style rules—this skill ensures that the model's limited attention is focused on high-value architectural context and essential project commands rather than redundant tasks better handled by deterministic tooling.

Key Features

Guides progressive disclosure by linking to external documentation for complex rules
3 GitHub stars
Manages instruction budgets to prevent model distraction and improve reliability
Identifies and eliminates the 'Claude as Linter' anti-pattern for better performance
Implements the WHAT/WHY/HOW framework for concise project-level context
Provides standardized templates for both user-level and project-level configurations

Use Cases

Initializing a new project with a high-performance CLAUDE.md file
Establishing global developer preferences through optimized user-level configs
Auditing bloated configuration files to improve AI instruction adherence

About

Key Features

Guides progressive disclosure by linking to external documentation for complex rules
3 GitHub stars
Manages instruction budgets to prevent model distraction and improve reliability
Identifies and eliminates the 'Claude as Linter' anti-pattern for better performance
Implements the WHAT/WHY/HOW framework for concise project-level context
Provides standardized templates for both user-level and project-level configurations

Use Cases

Initializing a new project with a high-performance CLAUDE.md file
Establishing global developer preferences through optimized user-level configs
Auditing bloated configuration files to improve AI instruction adherence