Skip to content

Meta Prompting to generate AI instructions, agents, skills & prompts

Notifications You must be signed in to change notification settings

JBurlison/MetaPrompts

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 

Repository files navigation

MetaPrompts

This repo contains AI coding assistant customization files meant to be copied into your own repo, centered on the meta agent ai-builder. Works with GitHub Copilot, Claude Code, Codex, OpenCode, and other providers using the same folder structure.

Provider Folder Structure

These customization files work across multiple AI coding assistant providers. Copy to the appropriate folder for your environment:

Provider Base Folder
GitHub Copilot .github/
Claude Code .claude/
Codex .codex/
OpenCode .config/opencode/

The internal structure (agents/, skills/, prompts/, instructions/) is the same across all providers.

Use the meta agent in VS Code

  1. Copy the .github folder (or rename to your provider's folder) into your repo.
  2. Open your repo in VS Code.
  3. Open the Copilot Chat view.
  4. Select the agent ai-builder from the agent picker, or type @ai-builder in the chat.
  5. Ask for help creating or maintaining AI assistant customizations.

Example requests:

  • Create a new agent for database schema design.
  • Validate my agent and skill files.
  • Generate documentation for the current agents.
  • Help me understand what agents, prompts, and instructions are.

Meta agent capabilities

The ai-builder agent can:

  • Design and create agents, skills, prompts, and instructions.
  • Recommend the right customization type for a request (agent vs prompt vs instructions vs workflow).
  • Build multi-agent workflows with handoffs and review points (following Requirements → Due Diligence → Plan → Implement → Review pattern).
  • Create sub-agents with proper naming (*.subagent.agent.md) and user-invokable: false.
  • Explain the framework to users unfamiliar with agents, prompts, skills, and instructions.
  • Validate and troubleshoot customization files for format issues.
  • Analyze overlaps and redundancies across agents, skills, prompts, and instructions.
  • Generate documentation and usage guides for your customizations.

What to copy

Replace <provider> with your provider's folder (.github, .claude, .codex, .config/opencode):

  • <provider>/agents/ai-builder.md — the meta agent definition
  • <provider>/skills/ — reusable procedures referenced by agents
  • <provider>/prompts/ — reusable prompt templates
  • <provider>/instructions/ — auto-applied instructions by file pattern

Folder structure

<provider>/                          # .github/, .claude/, .codex/, etc.
├── agents/
│   ├── my-agent.agent.md            # User-facing agent
│   └── helper.subagent.agent.md     # Sub-agent (workflow component)
├── skills/
│   └── skill-name/
│       └── SKILL.md                 # Skill definition
├── prompts/
│   └── my-prompt.prompt.md          # Prompt template
└── instructions/
    └── coding.instructions.md       # Auto-applied instructions

Notes

  • Keep reusable procedures in skills and reference them from agents.
  • Add project-wide guidance in instructions so it applies automatically.
  • Use .subagent.agent.md naming and user-invokable: false for workflow sub-agents.
  • Commit changes to share the configuration across the team.

About

Meta Prompting to generate AI instructions, agents, skills & prompts

Resources

Stars

Watchers

Forks

Packages

No packages published