agent-instructions

Architect a canonical AGENTS.md with tool-specific shims for Claude, Copilot, Cursor, Gemini, and Codex.

2|Updated Apr 20, 2026
One-click install
npx skills add https://github.com/devantler-tech/agent-plugins --skill agent-instructions-devantler-tech
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: agent-instructions
Source: https://github.com/devantler-tech/agent-plugins/tree/main/plugins/agentic-engineering/skills/agent-instructions
Command: npx skills add https://github.com/devantler-tech/agent-plugins --skill agent-instructions-devantler-tech

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill prevents AI instruction files from drifting by establishing one canonical source of truth for a repository and wiring every supported tool to it cleanly.

Core Features & Use Cases

  • Canonical Guidance: Defines what belongs in AGENTS.md so durable, tool-neutral repo guidance lives in one place.
  • Tool-Specific Shims: Shows how Claude Code and Gemini can include the canonical file instead of duplicating it.
  • Copilot Subset Maintenance: Explains how to keep a concise, review-focused Copilot instruction file aligned with the canonical source.
  • Use Case: Use it when setting up a new repo for multiple AI coding tools or when fixing stale, contradictory instruction files.

Quick Start

Ask this skill to help you design or repair your repository’s agent instruction files so AGENTS.md becomes the canonical source and every tool-specific file stays synchronized.

Frequently Asked Questions about agent-instructions

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I keep AI coding assistant instructions from drifting across multiple tools?

To prevent AI coding assistant instructions from drifting, establish one canonical AGENTS.md file as a tool-neutral source of truth, then wire Claude Code, Cursor, and Gemini to it using thin include-based shims instead of duplicating configuration.

What is the best way to structure AGENTS.md for Claude Code, Copilot, and Gemini?

The best way to structure AGENTS.md for multiple AI coding tools is to maintain a canonical, tool-neutral file for primary guidance, use include-based shims for Claude Code and Gemini, and keep a concise Copilot subset aligned with the main source.

How do I synchronize Copilot instructions with a canonical AGENTS.md file?

You synchronize Copilot instructions with a canonical AGENTS.md file by maintaining a concise, review-focused Copilot instruction subset that stays strictly aligned with the durable, tool-neutral guidance defined in your primary configuration.

Does this approach work for both setting up new repositories and repairing stale agent instruction files?

Yes, this approach works for both setting up new repositories and repairing stale agent instruction files by applying a canonical AGENTS.md architecture to establish or restore synchronization across Claude, Copilot, Cursor, Gemini, and Codex.

When should I use include-based shims for repository AI guidance?

You should use include-based shims for repository AI guidance when you need to connect tools like Claude Code and Gemini to a canonical AGENTS.md file without duplicating content, ensuring updates propagate cleanly without configuration drift.

Why are my AI coding assistant instructions contradicting each other across different tools?

AI coding assistant instructions contradict each other across different tools because of configuration drift caused by duplicating guidance instead of using a canonical AGENTS.md source with thin include-based shims for synchronization.