A Fable of Codexes skill
A skill that makes Claude a conductor: it hands tasks to Codex workers in git branches and verifies each before merge.
- 56 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- Codex workers write code in parallel on branches of your repository and consume subscription limits or paid tokens.
- The conductor merges branches after its own verification, so review merges yourself.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/a-fable-of-codexesDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Codex workers write code in parallel on branches of your repository and consume subscription limits or paid tokens.
- The conductor merges branches after its own verification, so review merges yourself.
This is third-party code. Review the repository files before installing.
What it does
The `campaign-conductor` skill turns Claude Code into the head of a fleet of OpenAI Codex workers. Claude plans the work, launches Codex workers in parallel, each in its own git worktree, and verifies every branch before merge. The conductor itself writes no feature code. Its context stays free for planning and judgment because it reads worker reports rather than their transcripts. Workers have their own tiers: for implementation, for independent review of sibling diffs and for mechanical work. Claude agents on Opus 5.5 handle design and lead squads of workers, while Sonnet 5.5 agents survey the codebase. Campaign state lives in `docs/campaign-hq/` inside the repository, so the next session resumes from the same place. Preferences such as sending tests to Opus are saved in `preferences.md`.
Who it is for. Developers with Claude and Codex subscriptions running a large project who want implementation spread across several agents.
Good fit when
- A task splits into independent pieces that can be written in parallel on different branches.
- A migration runs across several sessions and state must live in the repository.
- An independent review of a diff by a second agent is needed before merge.
- An architecture question or a bake-off should go to a read-only reviewer.
Not a fit when
- The task is small and one agent copes without a fleet.
- Codex CLI is not installed or logged in: workers launch through it.
- Subscription limits are tight: waves of workers use them up fast.
Example request
Start a campaign: split the billing migration into phases, send implementation to Codex workers and show status from the plan.Limitations
Claude Code with Fable 5.1 or Opus 5.5 as the session model and OpenAI Codex CLI logged in with `codex login` are required. ChatGPT-plan login draws on plan limits, an API key is billed per token, so size waves to your limits. The skill uses the Agent and Workflow tools and sets model and effort on every dispatch. Deep consults have a separate maximum-effort model that runs only on explicit request.
How to disable. Delete the `campaign-conductor` folder from `~/.claude/skills/`. Keep or delete the `docs/campaign-hq/` directory and the git worktree branches by hand.
FAQ
How do I install the skill?
`npx skills add jvogan/a-fable-of-codexes --skill campaign-conductor`. By hand: clone the repository and copy the `skills/campaign-conductor` folder into `~/.claude/skills/`.
How do I start a campaign?
Tell Claude "start a campaign". It creates `docs/campaign-hq/`, sizes the plan to the project and starts handing tasks to workers.
Which commands does the conductor understand?
Plain phrases: add a phase, send tests to Opus, ask for a second opinion or show status. Persistent preferences are saved in `preferences.md`.
What else is needed besides Claude Code?
OpenAI Codex CLI: `npm install -g @openai/codex` or `brew install codex`, then `codex login`.
How are collisions between workers avoided?
Every writing worker gets its own git worktree, and the conductor verifies a branch before merge.
Discussion
No questions or reviews yet.
Related
SkillsA skill that makes the lead Claude model a foreman: it plans work, hands tasks to worker agents and verifies the result itself.
The /codex-workflows skill: Claude Code writes a script, dozens of Codex agents run it, and the run shows on a map
A Claude Code skill that hands implementation-heavy work to Codex CLI while Claude keeps planning, review and acceptance.
A skill that makes Claude a manager: it writes tickets, hands code to cheaper agents, verifies and asks another vendor to review.
An execution-discipline skill for Claude: staged planning, delegation to agents, failable checks and a double-check before delivery
One agent launches others in the background: Claude Code, Codex, Copilot CLI, Grok Build and Pi, with a model chosen per stage.
README in short
The README covers the fleet composition and model roles, one-command install and how to start a campaign. Requirements, Codex configuration options and verification principles before merge follow.