agent-os

name	agent-os
description	Initialize or synchronize "Agent-First" infrastructure (AGENTS.md, CONTEXT.md) in any repository. Use when you land in a new or undocumented codebase and need to make it instantly navigable for AI agents.

Skill: agent-os

Turn any repository into a high-density "Agent-Ready" workspace.

Usage

Initialize a new repo

./scripts/init.sh

Sync/Update context

./scripts/sync.sh

Guardrails

• READ: Root directory structure and package.json, requirements.txt, etc.
• WRITE: AGENTS.md, CONTEXT.md, and .workflow/ directory.
• NEVER: Overwrite a custom AGENTS.md without creating a backup.
• NEVER: Read .env or .secrets files.

Workflow

1. Init (Bootstrap)

• Runs scripts/discover.sh to identify the tech stack.
• Drops AGENTS.md with the mandatory ⚡️ AGENT BOOTSTRAP header.
• Drops CONTEXT.md with structured sections for Domain and Tech Stack.
• Creates a git-ignored .workflow/ folder for task tracking.

2. Sync (Anti-Staleness)

• Scans for new dependencies or structural changes.
• Compares findings against CONTEXT.md.
• Proposes updates to keep context dense and accurate.

Output contract

End with:

• Status: (Initialized / Synced)
• Detected Stack: (e.g. Node/TypeScript/Jest)
• Files Created/Updated: (List)
• Next step: Agent should perform a "Firmware Load" by reading the new docs.