GitHub, MCP, skills

Developer tools for AI-agent workflows.

MPG ONE builds practical tooling around the way modern teams use AI coding agents: repository checks, integration wrappers, reusable skills, and lightweight command-line systems.

mpg-one

RepoGuard

A static repository risk scanner for AI coding agents such as Codex, Claude Code, Cursor, Gemini CLI, and similar tools.

CLI availableStatus
PythonLanguage
2GitHub stars
Public tooling

GitHub repositories

Useful products and utilities built by MPG ONE, starting with RepoGuard for pre-agent repository scanning.

Integration layer

MCP servers

Wrappers that make MPG ONE tools callable from AI clients and operational workflows.

Agent workflow

Codex skills

Reusable instructions and commands that help AI coding agents use tools correctly across new projects.

RepoGuard

Scan before you agent.

RepoGuard checks a repository before an AI coding agent starts reading files, following instructions, installing dependencies, or editing code. It performs static analysis only, so it does not run repository code or setup scripts.

Open GitHub repo
01 Agent-targeted prompt injection
02 Dangerous install scripts
03 Sensitive-file access with network behavior
04 Suspicious GitHub Actions workflows
05 Obfuscated JavaScript and miner indicators

Integration roadmap

Built once, usable by people, CI, and agents.

CLI-first products that work without a heavy dashboard

JSON and SARIF output for agents, CI, and security pipelines

MCP wrappers for Claude, Cursor, Codex-compatible clients, and other MCP-capable tools

Codex skill wrappers that run checks before agents work inside unknown repositories