Architecture
Repository Layout
| Path | Role |
|---|---|
SKILL.md | Trigger description, workflow guidance, and usage expectations for Codex |
agents/openai.yaml | UI metadata for skill discovery surfaces |
scripts/lib/github-project.mjs | Shared argument parsing, gh wrappers, and utility helpers |
scripts/*.mjs | Focused commands for schema inspection and project item updates |
references/ | Reusable notes for command patterns and field IDs |
docs/ | Public-facing English and Japanese docs built with VitePress |
Data Flow
- The skill or a developer chooses a project owner and project number.
- Helper scripts call
gh projectand request JSON output. - Shared helpers normalize that data into readable JSON or a targeted update call.
- The repo documents both the raw
ghpath and the Node helper path so the workflow stays portable.
Design Choices
Keep The Helpers Small
Each script owns one narrow operation. That keeps failures easier to debug and makes it simpler to compose repeatable automations.
Prefer Live Discovery
Project IDs, field IDs, and single-select options are discovered from GitHub instead of hard-coded into the skill.
Preserve Cross-Platform Behavior
Project automation is intentionally Node-based so the repo does not depend on PowerShell, Bash-specific functions, or local shell aliases.
Protect The Public Surface
Repository QA checks keep README, docs, skill metadata, and key file references aligned as the repo evolves.