菜单
Run a Codex fan-out (review panel, adversarial verify, implement-then-verify) with the native runner, with no Claude agents in the loop. Use for "/codex-bridge:workflow <goal>".
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/collisionengineers/claude-using-codex-skill --skill workflow命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
This skill should be used whenever the user asks to "use codex", "ask codex", "delegate to codex", "second opinion from codex/gpt", "have codex review", "run this in codex", "codex workflow", "let gpt implement", or wants OpenAI Codex involved in review, planning, implementation, research, or a Workflow, and when deciding whether a task is better handed to Codex than done in Claude.
Get a read-only second opinion from Codex on a question, design, hypothesis, or diagnosis. Use for "/codex-bridge:consult <question>".
Check that Codex is installed, authenticated, and reachable through app-server; list models, effort tiers, presets, and the shared server state. Use for "/codex-bridge:doctor" and when Codex runs fail.
| name | workflow |
| description | Run a Codex fan-out (review panel, adversarial verify, implement-then-verify) with the native runner, with no Claude agents in the loop. Use for "/codex-bridge:workflow <goal>". |
| argument-hint | <goal, e.g. "review my uncommitted changes from 3 angles" or "implement task.md and verify"> |
| disable-model-invocation | true |
| allowed-tools | Bash(node:*), Bash(git:*), Read, Write, AskUserQuestion |
Run a Codex fan-out with scripts/codex-workflow.mjs. Every stage is a Codex turn: nothing relays
output, a stage may run for as long as Codex needs, and finished read-only calls are cached for
--resume. See ${CLAUDE_PLUGIN_ROOT}/skills/using-codex/references/native-runner.md for the API.
workflows/review-panel.mjs ({lenses: [{name, taskFile}], votes?})workflows/adversarial.mjs ({scope, max?, votes?})workflows/implement-verify.mjs ({taskFile, maxRounds?, cwd?})codex-bridge.mjs doctor --json (run_dir) or the project's .claude/ folder. Each file is a short brief: scope, angle, what counts as a finding. Pass the paths through --args.node "${CLAUDE_PLUGIN_ROOT}/scripts/codex-workflow.mjs" run "${CLAUDE_PLUGIN_ROOT}/workflows/<script>.mjs" --args @args.json
run_in_background: true. status --last is safe to call at any time.result from the summary (or read result_file). A result with ok: false names the stage that failed; do not present it as a clean run. Read out_file paths only for items the user wants detail on.When your own judgement belongs between Codex stages, run one workflow, read its result.json,
decide, and launch the next. That costs one Bash call per stage and no relaying agent.