agent-scripts: a centralized hub for shared agent instructions and portable workflow skills across local workspaces

What it solves

This project provides a centralized, portable system for managing shared instructions, reusable skills, and helper scripts for AI agents (such as Codex and Claude) across multiple local workspaces.

How it works

The system uses a directory structure to store shared rules in AGENTS.MD and operational skills in the skills/ folder. These are integrated into AI agents via symlinks to the user's global configuration directories (e.g., ~/.codex/ or ~/.claude/). It includes a set of dependency-light helper scripts to automate tasks like skill validation, committing changes, and browser automation via Chrome DevTools.

Who it’s for

Developers who use AI agents in their local development environment and want a single source of truth for agent behavior, shared skills, and portable utility scripts across different projects.

Highlights

• Centralized Agent Rules: Uses a pointer-style AGENTS.MD to maintain global rules while allowing repo-specific overrides.
• Modular Skill System: Implements a routing layer where skills are defined in Markdown files with YAML front matter for easy agent discovery.
• Portable Helpers: Includes specialized scripts for browser automation, skill validation, and controlled git commits.
• Cross-Repo Syncing: Supports symlinking repo-owned skills from other projects into a central hub for unified access.