The Codex Chess Playing Agent — Development skill for Claude Code
Autonomous macOS Chess Agent with Computer Use - powered by GLM 5.6 Luna & Agent Skills standard.
How to install The Codex Chess Playing Agent
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open romangalaxys10-spec/The-Codex-Chess-Playing-Agent and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What The Codex Chess Playing Agent does
Autonomous macOS Chess Agent with Computer Use - powered by GLM 5.6 Luna & Agent Skills standard.
Alternatives in Development
- Ralphy — My Ralph Wiggum setup, an autonomous bash script that runs Claude Code, Codex, OpenCode, Cursor agent, Qwen & 2.8k ★
- Codenotch — A macOS app that pins usage limits from Claude Code, Cursor, Codex, and Antigravity to a screen edge 1.3k ★
- Chops — 1,000+ macOS app to browse, edit, and manage skills across Claude Code, Cursor, Codex, Windsurf, and Amp 1.1k ★
README
♟️ The Codex Chess Playing Agent 🚀
**Autonomous Computer-Use Chess Agent for macOS** — Pioneered with **GPT 5.6 LUNA** (Codex), then upgraded by **GLM 5.3 via ZCode** to fully control the desktop chess game, piece by piece, via macOS Accessibility.
[](LICENSE) [](https://apple.com) [](#)
🌟 Origin Story: GPT 5.6 LUNA Creates the Codex Version
The original breakthrough came from **GPT 5.6 LUNA** running in Codex: give an LLM direct "Computer Use" powers over macOS and challenge the native Apple Chess engine. LUNA pioneered the core insight of this repo — **skip OCR entirely and sniff the game state** straight out of Chess.app's autosaved document (`~/Library/Containers/com.apple.Chess/...`), reconstruct the board, and drive the desktop via AppleScript. That architecture — state sniffing + desktop control, packaged as a portable Agent Skill — is LUNA's contribution, and it works across Codex, Claude Code, Antigravity, ZCode, Trae, and Cursor.
🔥 The Upgrade: GLM 5.3 (via ZCode) Makes It Actually Move Pieces
On 2026-08-23, **GLM 5.3 running inside ZCode** loaded this skill from GitHub and pointed it at a **live, already-open Chess.app game** on the user's Mac desktop. What followed was a genuine agentic debugging session — three dead ends, one accidental discovery, and a final executor that plays moves deterministically with zero vision and zero mouse math.
Dead end #1 — the parser was reading the wrong file format
The shipped parser looked for a `` XML tag inside the `.game` file. Apple Chess actually persists a **plist** whose keys are `Moves` (newline-separated coordinate moves like `d2d3`) and `Position` (a live FEN string). GLM 5.3 rewrote the reader
Related Skills
Quota Peek
所有大模型额度,一眼看清 — 本地优先的 macOS 菜单栏工具,查看 Codex/Claude/Kimi/GLM 订阅额度
TokenLink
Native macOS menu-bar monitor for Codex, Claude, Kimi, MiniMax, and GLM coding plans, with optional M5Stack St
Opentrapp
A safer way to run autonomous CLI agents on your own computer. Open-source, MIT, community-driven.
Ralph Go
Start the autonomous Ralph loop. Usage: /ralph-go [cautious standard trusting]
Story Lifecycle
Standard Jira story lifecycle for autonomous agent
KyZN
Autonomous code improvement CLI powered by Claude Code
Related Agents
Computer Use Operator
Desktop automation specialist. Controls native macOS apps via Computer Use MCP. Handles Finder, System Setting
GPT 5.6 Luna
General-purpose agent that runs on GPT-5.6-Luna (routed through the shunt gateway to the ChatGPT/Codex subscri
Fable Reviewer
codex-luna-max의 구현 결과를 PLAN.md 대비로 검수한다. diff를 읽고 검증 명령을 실제로 돌려서 PASS/FAIL 판정과 지적사항을 REVIEW.md에 쓴다. 코드는 고치지 않는