I turned OpenAI Symphony into a one-command local workflow for any repo

OpenAI Symphony is powerful, but for most developers the first problem is not capability.

It’s setup friction.

You still need to:

• wire Linear correctly
• create a usable workflow file
• bootstrap local scripts into the repo
• make the setup portable across machines
• restart it cleanly every time you reopen Codex

I got tired of doing that manually, so I packaged the setup into a reusable public skill: Codex Symphony

What it gives you

Codex Symphony bootstraps a local Symphony + Linear workflow into any Git repository.

It installs:

• WORKFLOW.symphony.md
• scripts/symphony/start-local.sh
• scripts/symphony/start-background.sh
• scripts/symphony/status.sh
• codex-symphony wrapper command

The result is a practical local flow:

• Linear = queue
• Codex = executor
• Symphony = dispatcher

Install

You can install it from the broader Citedy skills package:

  npx @citedy/skills install symphony

Package:
https://www.npmjs.com/package/@citedy/skills

Or install the standalone skill:

  npx openskills install Citedy/codex-symphony

Standalone repo:
https://github.com/Citedy/codex-symphony

Required env

Add:

• LINEAR_API_KEY
• LINEAR_PROJECT_SLUG
• SOURCE_REPO_URL
• SYMPHONY_WORKSPACE_ROOT
• optional GH_TOKEN

Start

Inside your target repo:

/codex-symphony

Or after bootstrap:

codex-symphony

PS

The real pain was not “can Symphony do useful work?”
It was “can I get from zero to a repeatable local workflow without rebuilding the same setup every time?”

This package is my answer to that.

If you’re already using Linear and Codex, it turns Symphony into something much easier to adopt in a real repository.