Use Codex CLI as a sub-agent inside Claude Code skills.
Claude excels at conversation and alignment. Codex excels at deep reasoning and code generation. This bridge lets you combine both — Claude orchestrates, Codex executes.
claude plugin install github:zongmin-yu/claude-codex-bridge
Then use /codex in any Claude Code conversation:
/codex investigate why the auth tests are flaky in src/auth/
Every Codex-powered skill follows three steps:
Turn the user's request into a self-contained instruction for Codex. Point to files by absolute path — don't paste contents.
codex exec --full-auto --sandbox read-only \
-C "/path/to/repo" \
"Your prompt here" 2>/dev/null
Pick a sandbox level:
| Level | Can read | Can write | Has network | Use for |
|---|---|---|---|---|
read-only | Yes | No | No | Research, review |
network-only | Yes | Yes | No | Coding |
full | Yes | Yes | Yes | Tasks needing network |
Check that Codex produced what was asked. If something is missing, resume:
echo "Also check the edge case for empty input" | codex exec resume <session-id> 2>/dev/null
That's it. The /codex skill included in this plugin implements exactly this pattern.
The real power is creating specialized skills for your workflow. Copy examples/template.md to skills/<name>/SKILL.md and fill in:
See examples/ for reference implementations:
research.md — investigate codebasescoding.md — implement featuresreview.md — code reviewEvery codex exec prints a session ID. Save it to continue a conversation:
echo "Follow-up question" | codex exec resume <session-id> 2>/dev/null
For complex tasks, increase reasoning:
codex exec --full-auto --sandbox read-only \
--config model_reasoning_effort="high" \
-C "/path/to/repo" "Your prompt" 2>/dev/null
Point Codex at any directory. One skill can work across multiple repos by dispatching separate codex exec calls with different -C paths.
npm install -g @openai/codex)OPENAI_API_KEY set in your environmentMIT
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Research operators for paper triage, reference expansion, and citation tracing via Semantic Scholar API
Move active Claude Code workflows between machines without losing the thread
npx claudepluginhub zongmin-yu/claude-codex-bridgeOne plugin to bridge and delegate across Claude Code, Codex CLI, and Antigravity CLI (agy) — single-source AGENTS.md, shared skills, mirrored hooks and MCP servers, and bidirectional delegation.
Codex CLI orchestration from Claude Code: /codex:ask read-only delegation, /codex:exec one-shot runs with JSONL monitoring, /codex:auto full issue lifecycle delegation, /codex:status readiness checks.
Coordinate reusable monitored OpenAI Codex agents from Claude Code.
Cross-agent review workflow: Claude implements, Codex reviews
Delegate coding tasks to the local OpenAI Codex CLI agent using ChatGPT Pro OAuth flat-rate subscription. Wraps `codex exec` and `codex review` for code generation, refactoring, and review without per-token API charges. Use when the user wants to run coding tasks through Codex, generate or refactor code, or run code reviews via the local Codex CLI.
Ultra-compressed communication mode. Cuts 65% of output tokens (measured) while keeping full technical accuracy by speaking like a caveman.