{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreidhfp5joauut77ulb7xackza72kb7erswcq5mjml4mpmkt4imk2vi",
    "uri": "at://did:plc:2u26gaflouttm3uj6jkgroyz/app.bsky.feed.post/3me2hntzan5n2"
  },
  "coverImage": {
    "$type": "blob",
    "ref": {
      "$link": "bafkreib6bdytotc7gmiibyu52cyliozgzqceshpgj2oabtutoub7uyedti"
    },
    "mimeType": "image/jpeg",
    "size": 68300
  },
  "description": "Up until recently, getting Claude to collaborate with other agents was…functional, but clunky. Recently, I started using a skill that allows me to just call OpenAI /codex from Claude.",
  "path": "/til/2026/02/03/how-i-got-claude-to-consult-codex-and-when-i-use-it/",
  "publishedAt": "2026-02-03T00:11:03.000Z",
  "site": "https://danielraffel.me",
  "tags": [
    "Install the /codex skill",
    "Install the /claude skill"
  ],
  "textContent": "> **TL;DR** If you want **Claude to get a second opinion from Codex and Codex to get a second opinion from Claude** , these skills make it easy.\n>\n> - Install the /codex skill inside Claude → enables `/codex` in Claude\n> - Install the /claude skill inside Codex → enables `/claude` in Codex\n>\n> Unless you use a shared top-level skill folder then put each skill in the correct `skills/` directory for that tool, restart, run `/skills` to confirm.\n\n* * *\n\n# What This Solves\n\nThe goal is simple: Get a second opinion without ceremony.\n\nNo artifact juggling.\nNo copying files back and forth.\nNo breaking flow.\n\nJust use a skill, get a response, continue.\n\nOne agent stays in charge. The other is consulted. The lead integrates the result.\n\n_Example real world prompt in Claude:_\n\n\n    Please review the original source using RepoPrompt to ground the analysis in the actual codebase and consult Apple documentation via Sosumi MCP. After forming a plan, review it with /codex ask it to also use RepoPrompt to ground the analysis and consult Apple documentation via Sosumi MCP, refine it, and produce an updated v2 document with revised phases and align on next steps.\n\n* * *\n\n# How It Works\n\nTwo skills. Two agents.\n\n  * Claude can call `/codex`\n  * Codex can call `/claude`\n\n\n\nWhichever agent you’re actively using remains primary. The other is consulted for a scoped task and returns output. You decide how to use that response in your prompt. _I usually have them consult and align before moving forward._\n\n* * *\n\n## Setup\n\nThe following approaches use sparse checkouts so you only download the exact skill folder you need rather than the entire repo.\n\n## Install `/codex` in Claude\n\n### 1. Pull only the `codex` skill\n\n\n    git clone --depth 1 --filter=blob:none --sparse https://github.com/danielraffel/generous-corp-marketplace.git\n    cd generous-corp-marketplace\n    git sparse-checkout set skills/codex\n\n### **2. Create Claude’s skills directory (if missing)**\n\n\n    mkdir -p ~/.claude/skills\n\n### **3. Move the skill into place**\n\n\n    mv skills/codex ~/.claude/skills/\n\n### **4. Restart Claude**\n\nVerify installation:\n\n\n    /skills\n\nYou should now see /codex.\n\n* * *\n\n## Install `/claude` in Codex\n\n### **1. Pull only the claude skill at the pinned commit**\n\n\n    git clone --filter=blob:none --sparse https://github.com/danielraffel/generous-corp-marketplace.git\n    cd generous-corp-marketplace\n    git sparse-checkout set skills/claude\n    git checkout cd7d30498caa7114036e18687243fde4ff081083\n\n### **2. Create Codex’s skills directory (if missing)**\n\n\n    mkdir -p ~/.codex/skills\n\n### **3. Move the skill into place**\n\n\n    mv skills/claude ~/.codex/skills/\n\n### **4. Restart Codex**\n\nVerify installation:\n\n\n    /skills\n\nYou should now see /claude.\n\n* * *\n\n## **Final Folder Structure**\n\nClaude:\n\n\n    ~/.claude/skills/codex/\n      agents/\n      SKILL.md\n\nCodex:\n\n\n    ~/.codex/skills/claude/\n      references/\n      SKILL.md\n\n* * *\n\n## **What You Now Have**\n\n  * Claude can consult Codex with `/codex`\n  * Codex can consult Claude with `/claude`\n\n\n\nClean second opinions. No artifact passing. No context juggling.\n\n* * *\n\n## Ralph Loop / Parallel Optimization\n\nIf you’re running a ralph-loop style workflow or keeping a CLAUDE.md you can explicitly allow delegation for parallelizable tasks.\n\nSomewhere I stumbled on someone suggesting using it in a `ralph-loop` or a `CLAUDE.md` and they shared this snippet that leverages the `/codex` skill:\n\n\n    CODEX DELEGATION (OPTIONAL):\n    - Use /codex <task> or codex exec --full-auto <task> for parallel work.\n    - Good candidates for Codex delegation:\n      - Writing tests for code you just wrote\n      - Implementing a component while you work on another\n      - Code review of completed work items\n    - Do NOT delegate to Codex when:\n      - The task depends on something you're currently building\n      - Multiple agents would edit the same file\n      - The task requires your current conversation context\n    - When delegating, run Codex in background and continue your work.\n    - Check Codex output before marking work item complete.\n\n### Working Ralph Loop Output",
  "title": "How I Got Claude and Codex to Consult Each Other via Skills (and When I Use Them)",
  "updatedAt": "2026-03-09T19:53:22.751Z"
}