Godogen Source Repo banner
htdt htdt

Godogen Source Repo

Development community intermediate

Description

This repository contains two source trees for the same project: - `claude/` — Claude Code source tree - `codex/` — Codex source tree This is not a published game repo. The files at the repo root descr

Installation

This entry records only its repository, not the path inside it, so there is no exact command to give. Open the source below and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

Repository README

This is the README for htdt/godogen, shared by 8 entries in this directory. It describes the repository, not this entry specifically.

Godogen Source Repo

This repository contains two source trees for the same project:

  • claude/ — Claude Code source tree
  • codex/ — Codex source tree

This is not a published game repo. The files at the repo root describe how to work on the source trees themselves.

Source vs Runtime

  • claude/game.md is the template that becomes CLAUDE.md in a published Claude game repo.
  • codex/game.md is the template that becomes AGENTS.md in a published Codex game repo.
  • claude/publish.sh publishes runtime skills into .claude/skills/.
  • codex/publish.sh publishes runtime skills into .agents/skills/.
  • Root docs such as README.md and setup.md describe the combined repo and shared workstation setup.

Skills

Both source trees carry the same three skills:

  • godogen — orchestrator + inline task execution + all pipeline stages
  • godot-api — Godot class API lookup
  • visual-qa — visual quality assurance via multimodal review

Editing Rules

  • User must specify which is the target source tree.
  • Variant-specific work stays in the matching subtree: claude/ or codex/.
  • Shared docs and shared repo policy live at the repo root.
  • Do not align the behavior across both variants unless asked.
  • Do not create or maintain .claude/skills/ or .agents/skills/ in this source repo.
  • If you change a Codex skill's user-facing purpose, also update its agents/openai.yaml metadata.
  • When writing skills: don't give obvious guidance. The agent is a highly capable LLM — handholding only pollutes the context.