Can I use this with other AI assistants besides Claude?

Yes, while optimized for Claude Code, the AgentsMD standard is vendor-agnostic and supports any AI tool that utilizes project-level instruction files.

What is the difference between an Agent and a Skill in this framework?

Agents are single-responsibility workers focused on task execution, while Skills define the canonical patterns and best practices that Agents reference to perform tasks correctly.

How does this skill handle token management?

It enforces a strict target of 500 tokens (with a 1,000 token maximum) per file to ensure instructions remain focused, efficient, and easy for the AI to process.

What is the primary benefit of the AgentsMD architecture?

It provides a single source of truth for AI instructions, using symlinks to prevent logic duplication across different AI vendor configurations like .claude, .copilot, and .gemini.

AgentsMD Authoring Standards

Name: AgentsMD Authoring Standards
Author: JacobPEvans

byJacobPEvans

•

Productivity & Workflow

Standardizes the creation and management of modular AI instructions, agents, and skills using the AgentsMD framework.

This skill establishes a canonical architecture for AI instructions by centralizing logic in an 'agentsmd' directory and using symlinks for vendor-specific configurations like Claude, Copilot, and Gemini. It enforces a two-tier system—distinguishing between task-executing Agents and pattern-defining Skills—to ensure instructions are DRY, modular, and optimized for token limits. By following these standards, developers can create highly maintainable, single-purpose AI components that improve consistency and performance across different AI coding environments.

Key Features

01Centralized 'agentsmd' directory as a single source of truth across all AI vendors.

02Vendor-agnostic symlinking strategy to prevent instruction duplication.

03Strict token management with a 500-1,000 token target per file for better model focus.

04Two-tier architecture separating reusable pattern Skills from task-specific Agents.

05Standardized naming conventions and frontmatter templates for modularity.

062 GitHub stars

Use Cases

01Scaling project-specific rules without exceeding model context or token limits.

02Ensuring consistency when managing multiple AI assistant configurations simultaneously.

03Organizing complex AI project instructions into manageable, reusable components.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jacobpevans/claude-code-plugins agentsmd-authoring

For use in Claude.ai and ChatGPT

Key Features

01Centralized 'agentsmd' directory as a single source of truth across all AI vendors.

02Vendor-agnostic symlinking strategy to prevent instruction duplication.

03Strict token management with a 500-1,000 token target per file for better model focus.

04Two-tier architecture separating reusable pattern Skills from task-specific Agents.

05Standardized naming conventions and frontmatter templates for modularity.

062 GitHub stars

Use Cases

01Scaling project-specific rules without exceeding model context or token limits.

02Ensuring consistency when managing multiple AI assistant configurations simultaneously.

03Organizing complex AI project instructions into manageable, reusable components.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jacobpevans/claude-code-plugins agentsmd-authoring

For use in Claude.ai and ChatGPT