Claudex Loop skill
Skills for cross-provider work between Claude Code and Codex: one agent drafts the plan, the other reviews it, then a build and an independent code inspection f
- 3,401 ★ on the repository
- Last commit
Extended access
Why this level
- Launches Claude and Codex CLI subprocesses
- In the build phase it edits project code and runs proof commands
Install
- Works in
Manual install
/plugin marketplace add chaseai-yt/claudex-loop
/plugin install claudex-loop@claudex-loopAfter install the claudex-loop, claudex-route, codex-review and codex-build commands are available.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Launches Claude and Codex CLI subprocesses
- In the build phase it edits project code and runs proof commands
This is third-party code. Review the repository files before installing.
What it does
The set gives the agent a workflow where neither the plan nor the code is graded by whoever produced it. The current session handles recon, requirements and a plan with checkable acceptance criteria, while the other provider reviews the plan on the merits and returns evidence-backed findings until it reaches a verdict. Once authorized, the chosen builder implements the plan, runs proof checks, and a fresh session of the provider that did not write the code performs the final inspection. Decisions and models are recorded in a separate log. The set holds four skills: claudex-route for model choice and one-off delegation, claudex-loop for the full cycle, and the compatibility commands codex-review and codex-build.
Who it is for. For developers who run both Claude Code and Codex and want an independent review of the plan and the code.
Good fit when
- You need to harden a plan before coding and get a review from a second model
- You want the code inspected by an agent other than the one that wrote it
- You need to pick a model for a task or make a one-off handoff
Not a fit when
- The task is trivial and a second-model review only slows things down
- Only one of the two CLIs is installed and configured
Example request
claudex this feature, mode=review, plan=docs/plan.md, rounds=3Limitations
The full cycle needs both CLIs installed and authenticated, plus Python 3.10 or newer for the shared adapter. No separate API keys or pip packages are required, but the Claude and OpenAI accounts are unavailable from Russia without a VPN. The skill does not remove user control: a review request does not authorize a build, and edit authorization is granted separately.
How to disable. Remove the claudex-loop plugin via /plugin, or delete the copied skill folders from ~/.claude/skills/ and ~/.agents/skills/.
FAQ
How does claudex-route differ from claudex-loop?
Route recommends a model and a role and can perform one handoff; it is self-contained. Loop is the full cycle with plan review, build and independent inspection.
Do I have to install both CLIs?
For the full cycle, yes, since review and inspection always go to the other provider. claudex-route can be installed on its own and used without the second CLI until a handoff is needed.
Discussion
No questions or reviews yet.
Related
SkillsA Claude Code skill that hands implementation-heavy work to Codex CLI while Claude keeps planning, review and acceptance.
A Claude Code skill with 21 roles and a stage graph with code-enforced quality gates for build, review, test and audit.
The /codex-workflows skill: Claude Code writes a script, dozens of Codex agents run it, and the run shows on a map
A Codex skill for an autonomous loop: Claude Code computes, Codex reviews hourly, critiques and sends the next prompt.
A planner for Claude Code, Codex and OpenCode: one goal becomes a task plan and independent tasks run in parallel
Claude Code supervises while Codex, opencode and pi work on your quota: delegated edits, independent review, model discussions
README in short
The README describes a repository of four skills for joint Claude Code and Codex work. The core idea: one provider reviews the plan, another builds, and the provider that did not write the code runs the final inspection, with all decisions written to a log. Behavior is driven by arguments: mode, plan file, builder, round count, models and per-role effort. It installs as a Claude Code plugin or by copying skill folders for Codex; the full cycle needs both CLIs and Python 3.10+. MIT licensed.
README badge
Original SKILL.md
--- name: claudex-loop description: "Harden a plan with independent Claude/Codex review, then optionally build and cross-inspect it. Start in either Claude Code or Codex: the host plans and the other provider reviews. Use for claudex this plan, claudex-loop, or the legacy crucible trigger; not for trivial edits." --- # Claudex Loop The current conversation owns requirements, planning and coordination. The other provider reviews the plan. Either provider can build; the provider that did not build inspects the final code in a fresh session. ## Resolve roles once Identify the actual host from your runtime, not PATH, installed skills, model-name guesses, or the repository. Both CLIs may be installed. Use `host=claude` in Claude Code and `host=codex` in Codex. If the runtime identity is unavailable, ask which host the user is using once. | Host | Requirements and plan | Plan reviewer | Default builder | Final inspector | |---|---|---|---|---| | Claude Code | Current Claude session | Codex | Claude | Fresh Codex session | | Codex | Current Codex session | Claude | Codex | Fresh Claude session | Honor `builder=claude|codex`. The inspector is always the other provider. The host remains coordinator even when the other provider builds.