← Back to Skills Marketplace
OpenClaw Claude Codex Workflow
by
Joshua Wang
· GitHub ↗
· v0.1.0
· MIT-0
165
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install openclaw-claude-codex-workflow
Description
在 OpenClaw 中需要 Claude Code 负责编排/评审、Codex 实现后端、Gemini 实现前端的多模型流水线时启用；尤其适用于需要 OPSX/spec 对齐、分阶段 artifacts、以及 claude→codex→gemini→claude 闭环交付的任务。
Usage Guidance
This skill appears to implement the multi-model orchestration it advertises, but its metadata fails to list the real runtime requirements. Before installing or running it:
- Do not run it in a production or privileged repository until you confirm its behavior. Prefer a disposable clone.
- Confirm you trust the claude/codex/gemini CLIs the script will invoke; those CLIs will receive prompts and context files and may execute commands.
- Provide API keys only if you understand the CLIs' security model; avoid placing secrets in global shell profiles if possible (use a dedicated process env or a secrets manager).
- Change the workspace from /root/.openclaw/workspace to a non-root workspace path to avoid requiring elevated privileges.
- Use --dry-run and review generated .claude/plan.md and any commands before letting the orchestrator execute them (set DRY_RUN=1 or pass --dry-run).
- Ask the publisher to update registry metadata to list required binaries and required environment variables (CLAUDE_API_KEY, CODEX_API_KEY, GOOGLE_API_KEY, and the required CLIs) so you can perform a complete risk assessment.
If you want, I can extract a checklist of the exact commands and env vars the skill expects so you can validate them locally before running.
Capability Analysis
Type: OpenClaw Skill
Name: openclaw-claude-codex-workflow
Version: 0.1.0
The skill bundle provides a legitimate orchestration framework for a multi-model development pipeline involving Claude, Codex, and Gemini. The primary logic resides in `scripts/ccg_orchestrator.sh`, which safely uses Bash arrays to wrap official CLI commands without evidence of shell injection or unauthorized data exfiltration. While the workflow requires sensitive API keys (e.g., `CLAUDE_API_KEY`, `GOOGLE_API_KEY`), the instructions in `SKILL.md` and the supporting documentation in the `references/` directory are transparent, align with the stated purpose, and include safety constraints such as dry-run modes and git-cleanliness checks.
Capability Assessment
Purpose & Capability
The SKILL.md and script require claude, codex, gemini CLIs, git, jq, Node.js, and API keys (CLAUDE_API_KEY, CODEX_API_KEY, GOOGLE_API_KEY), and reference a default workspace (/root/.openclaw/workspace). However the registry metadata declares no required binaries, no required environment variables, and no config paths. This mismatch between claimed metadata and actual runtime requirements is incoherent and surprising to users.
Instruction Scope
The instructions confine model roles and prescribe writing artifacts to a .claude directory and to the git tree, and the orchestrator script only wraps the external CLIs (plan -> codex exec -> gemini run -> review). That scope is consistent with the stated purpose. Notes of concern: the guidance expects users to paste OPSX tickets and context into .claude/context.md (which may contain sensitive info), and the default workspace path targets /root which is atypical and implies elevated or global scope.
Install Mechanism
This is instruction-only with a small shell script — there is no download/install step in the registry. The contained script simply invokes external CLIs and performs git checks; nothing in the script pulls remote code or writes unexpected files outside the repo/worktree. Install risk is therefore low, but runtime still depends on external CLIs.
Credentials
The skill expects several credentials and CLI binaries (CLAUDE_API_KEY, CODEX_API_KEY, GOOGLE_API_KEY; CLAUDE_CLI/CODEX_CLI/GEMINI_CLI) yet the registry metadata declares none. Requiring multiple API keys is proportionate to a multi-model orchestration, but failing to declare them in metadata is a serious transparency gap. Also recommending exporting keys in shell profiles makes them available to any process that inherits the shell environment.
Persistence & Privilege
always:false and default autonomous invocation are unchanged — no elevated platform privileges requested. The skill asks to create/commit .claude artifacts in the repo and recommends a workspace under /root/.openclaw/workspace; that workspace choice increases required privilege and should be changed to a non-root path when possible. The skill does not modify other skills or global agent configs.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install openclaw-claude-codex-workflow - After installation, invoke the skill by name or use
/openclaw-claude-codex-workflow - Provide required inputs per the skill's parameter spec and get structured output
Version History
v0.1.0
Initial release
Metadata
Frequently Asked Questions
What is OpenClaw Claude Codex Workflow?
在 OpenClaw 中需要 Claude Code 负责编排/评审、Codex 实现后端、Gemini 实现前端的多模型流水线时启用；尤其适用于需要 OPSX/spec 对齐、分阶段 artifacts、以及 claude→codex→gemini→claude 闭环交付的任务。 It is an AI Agent Skill for Claude Code / OpenClaw, with 165 downloads so far.
How do I install OpenClaw Claude Codex Workflow?
Run "/install openclaw-claude-codex-workflow" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is OpenClaw Claude Codex Workflow free?
Yes, OpenClaw Claude Codex Workflow is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does OpenClaw Claude Codex Workflow support?
OpenClaw Claude Codex Workflow is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created OpenClaw Claude Codex Workflow?
It is built and maintained by Joshua Wang (@shennng); the current version is v0.1.0.
More Skills