XYZ Forge is a local-first engineering operations system. It coordinates several AI coding agents — Claude Code, Codex, agy (Google's Antigravity CLI), and others — working on the same repositories, and it carries the surrounding machinery an unattended agent run actually needs: specified work, collision-free lanes, verification gates, and receipts
This repository also configures its own agents. See what XYZ-forge tells them →
Skill Claude CodeCodex
Manage a durable local skill collection and global app symlinks through conversation. Use to import, update, list, remove, or deploy skill folders from repositories already on disk, configure deployment targets, or replace skills-sync-trinity. Not for remote downloads, publishing skills, or installing their runtime…
Skill Claude Code
Write copy in a specific human voice using a three-role, layered pipeline — a Cartographer that maps the voice from real samples, a Drafter that writes conditioned on verbatim exemplars, and a blind Adversarial Reader that never sees the voice map and grades against a planted control. Use when the user wants copy that…
Skill Codex
Interrupt a stalled AI work session and restore movement toward the original outcome. Use when the agent is overengineering, inventing machinery around the work, reopening settled decisions, exhausting review rounds without qualifying progress, polishing controls after the plan is ready, or reporting lots of activity…
Skill Claude Code
The 5-and-5 decision checksum for a plan, feature, or bug fix that is in scope: exactly five load-bearing things the artifact DOES — the decisions and behaviors, especially choices the agent made that nobody explicitly asked for — and five things it explicitly does NOT do, so unwelcome decisions surface before…
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: