Skip to content
Back to search

jarrodwatts-claude-delegator

MarketplaceWorks with Claude CodeGitHub

Delegate tasks to Codex and Gemini directly from within Claude Code.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/jarrodwatts-claude-delegator/badge.svg)](/listings/jarrodwatts-claude-delegator)

HTML

<a href="/listings/jarrodwatts-claude-delegator"><img src="/api/listings/jarrodwatts-claude-delegator/badge.svg" alt="findsafeskills safety rating" /></a>

Install

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

Inside a Claude Code instance, run the following commands:

**Step 1: Add the marketplace**

/plugin marketplace add jarrodwatts/claude-delegator

**Step 2: Install the plugin**

/plugin install claude-delegator

**Step 3: Run setup**

/claude-delegator:setup

Done! Claude now routes complex tasks to GPT experts automatically.

> **Note**: Requires [Codex CLI](https://github.com/openai/codex) or [Gemini CLI](https://github.com/google/gemini-cli). Setup guides you through installation.

---

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 32 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. {
  2. "name": "jarrodwatts-claude-delegator",
  3. "owner": {
  4. "name": "Jarrod Watts",
  5. "url": "https://github.com/jarrodwatts"
  6. },
  7. "plugins": [
  8. {
  9. "name": "claude-delegator",
  10. "source": "./",
  11. "description": "GPT expert subagents for Claude Code via Codex CLI. Five specialized experts: Architect, Plan Reviewer, Scope Analyst, Code Reviewer, Security Analyst.",
  12. "version": "1.0.0",
  13. "author": {
  14. "name": "Jarrod Watts",
  15. "url": "https://github.com/jarrodwatts"
  16. },
  17. "homepage": "https://github.com/jarrodwatts/claude-delegator",
  18. "repository": "https://github.com/jarrodwatts/claude-delegator",
  19. "license": "MIT",
  20. "keywords": [
  21. "delegation",
  22. "mcp",
  23. "codex",
  24. "gpt",
  25. "openai",
  26. "experts",
  27. "subagents",
  28. "orchestration"
  29. ]
  30. }
  31. ]
  32. }
README.md · README · 247 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. # Claude Delegator
  2. > [!WARNING]
  3. > **This project is archived and no longer maintained.** Codex CLI removed the `codex mcp-server` command this plugin relies on (deprecated in 0.149.1, gone after 0.153), so delegating to Codex no longer works with current Codex. The Gemini bridge is untested against current Gemini CLI releases.
  4. >
  5. > To use Codex from Claude Code, install OpenAI's official plugin, [openai/codex-plugin-cc](https://github.com/openai/codex-plugin-cc):
  6. >
  7. > ```
  8. > /plugin marketplace add openai/codex-plugin-cc
  9. > /plugin install codex@openai-codex
  10. > /codex:setup
  11. > ```
  12. >
  13. > Thanks to everyone who used and contributed to Claude Delegator.
  14. GPT expert subagents for Claude Code. Five specialists that can analyze AND implement—architecture, security, code review, and more.
  15. [![License](https://img.shields.io/github/license/jarrodwatts/claude-delegator?v=2)](LICENSE)
  16. [![Stars](https://img.shields.io/github/stars/jarrodwatts/claude-delegator?v=2)](https://github.com/jarrodwatts/claude-delegator/stargazers)
  17. ![Claude Delegator in action](claude-delegator.png)
  18. ## Install
  19. Inside a Claude Code instance, run the following commands:
  20. **Step 1: Add the marketplace**
  21. ```
  22. /plugin marketplace add jarrodwatts/claude-delegator
  23. ```
  24. **Step 2: Install the plugin**
  25. ```
  26. /plugin install claude-delegator
  27. ```
  28. **Step 3: Run setup**
  29. ```
  30. /claude-delegator:setup
  31. ```
  32. Done! Claude now routes complex tasks to GPT experts automatically.
  33. > **Note**: Requires [Codex CLI](https://github.com/openai/codex) or [Gemini CLI](https://github.com/google/gemini-cli). Setup guides you through installation.
  34. ---
  35. ## What is Claude Delegator?
  36. Claude gains a team of GPT and Gemini specialists via native MCP. Each expert has a distinct specialty and can advise OR implement.
  37. **Note:** You can use either provider (GPT or Gemini), or both. The plugin will automatically detect which one is configured and route tasks accordingly.
  38. | What You Get | Why It Matters |
  39. |--------------|----------------|
  40. | **5 domain experts** | Right specialist for each problem type |
  41. | **GPT or Gemini** | Use your preferred model provider |
  42. | **Dual mode** | Experts can analyze (read-only) or implement (write) |
  43. | **Auto-routing** | Claude detects when to delegate based on your request |
  44. | **Synthesized responses** | Claude interprets expert output, never raw passthrough |
  45. ### The Experts
  46. | Expert | What They Do | Example Triggers |
  47. |--------|--------------|------------------|
  48. | **Architect** | System design, tradeoffs, complex debugging | "How should I structure this?" / "What are the tradeoffs?" |
  49. | **Plan Reviewer** | Validate plans before you start | "Review this migration plan" / "Is this approach sound?" |
  50. | **Scope Analyst** | Catch ambiguities early | "What am I missing?" / "Clarify the scope" |
  51. | **Code Reviewer** | Find bugs, improve quality | "Review this PR" / "What's wrong with this?" |
  52. | **Security Analyst** | Vulnerabilities, threat modeling | "Is this secure?" / "Harden this endpoint" |
  53. ### When Experts Help Most
  54. - **Architecture decisions** — "Should I use Redis or in-memory caching?"
  55. - **Stuck debugging** — After 2+ failed attempts, get a fresh perspective
  56. - **Pre-implementation** — Validate your plan before writing code
  57. - **Security concerns** — "Is this auth flow safe?"
  58. - **Code quality** — Get a second opinion on your implementation
  59. ### When NOT to Use Experts
  60. - Simple file operations (Claude handles these directly)
  61. - First attempt at any fix (try yourself first)
  62. - Trivial questions (no need to delegate)
  63. ---
  64. ## How It Works
  65. ```
  66. You: "Is this authentication flow secure?"
  67. ↓
  68. Claude: [Detects security question → selects Security Analyst]
  69. ↓
  70. ┌───────────────────────────────┐
  71. │ mcp__codex__codex │
  72. │ (or mcp__gemini__gemini) │
  73. │ → Security Analyst prompt │
  74. │ → Expert analyzes your code │
  75. └───────────────────────────────┘
  76. ↓
  77. Claude: "Based on the analysis, I found 3 issues..."
  78. [Synthesizes response, applies judgment]
  79. ```
  80. **Key details:**
  81. - Each expert has a specialized system prompt (in `prompts/`)
  82. - Claude reads your request → picks the right expert → delegates via MCP (GPT or Gemini)
  83. - Responses are synthesized, not passed through raw
  84. - Experts can retry up to 3 times before escalating
  85. - Multi-turn conversations preserve context via `threadId` for chained tasks
  86. ### Multi-Turn Conversations
  87. For chained implementation steps, the expert preserves context across turns:
  88. ```
  89. Turn 1: mcp__*__* → returns threadId
  90. Turn 2: mcp__*__*-reply(threadId) → expert remembers turn 1
  91. Turn 3: mcp__*__*-reply(threadId) → expert remembers turns 1-2
  92. ```
  93. Use single-shot (`codex` or `gemini` only) for advisory tasks. Use multi-turn for implementation chains and retries.
  94. ---
  95. ## Configuration
  96. ### Operating Modes
  97. Every expert supports two modes based on the task:
  98. | Mode | Sandbox | Use When |
  99. |------|---------|----------|
  100. | **Advisory** | `read-only` | Analysis, recommendations, reviews |
  101. | **Implementation** | `workspace-write` | Making changes, fixing issues |
  102. Claude automatically selects the mode based on your request.
  103. ### Configuration Defaults
  104. Set global defaults in `~/.codex/config.toml` instead of passing parameters on every call:
  105. ```toml
  106. sandbox_mode = "workspace-write"
  107. approval_policy = "on-failure"
  108. ```
  109. Per-call parameters override these defaults. See [Codex CLI docs](https://github.com/openai/codex) for all config options.
  110. ### Manual MCP Setup
  111. If `/setup` doesn't work, register the MCP server(s) manually:
  112. ```bash
  113. # For Codex (GPT)
  114. # Idempotent: safe to rerun
  115. claude mcp remove codex >/dev/null 2>&1 || true
  116. claude mcp add --transport stdio --scope user codex -- codex -m gpt-5.3-codex mcp-server
  117. # For Gemini
  118. # Idempotent: safe to rerun
  119. claude mcp remove gemini >/dev/null 2>&1 || true
  120. claude mcp add --transport stdio --scope user gemini -- node ${CLAUDE_PLUGIN_ROOT}/server/gemini/index.js
  121. ```
  122. Verify with:
  123. ```bash
  124. claude mcp list
  125. printf '{"jsonrpc":"2.0","id":"health","method":"initialize","params":{}}\n' | node ${CLAUDE_PLUGIN_ROOT}/server/gemini/index.js
  126. ```
  127. ### Customizing Expert Prompts
  128. Expert prompts live in `prompts/`. Each follows the same structure:
  129. - Role definition and context
  130. - Advisory vs Implementation modes
  131. - Response format guidelines
  132. - When to invoke / when NOT to invoke
  133. Edit these to customize expert behavior for your workflow.
  134. ---
  135. ## Requirements
  136. You need at least one of the following providers configured:
  137. - **Codex CLI** (for GPT): `npm install -g @openai/codex`
  138. - **Gemini CLI** (for Gemini): `npm install -g @google/gemini-cli`
  139. **Authentication**:
  140. - Codex: run `codex login`
  141. - Gemini: run `gemini` once and complete the sign-in flow (or set `GOOGLE_API_KEY`)
  142. ---
  143. ## Commands
  144. | Command | Description |
  145. |---------|-------------|
  146. | `/claude-delegator:setup` | Configure MCP server and install rules |
  147. | `/claude-delegator:uninstall` | Remove MCP config and rules |
  148. ---
  149. ## Troubleshooting
  150. | Issue | Solution |
  151. |-------|----------|
  152. | MCP server not found | Restart Claude Code after setup |
  153. | Provider not authenticated | Codex: run `codex login`. Gemini: run `gemini` once to complete sign-in (or set `GOOGLE_API_KEY`) |
  154. | Tool not appearing | Run `claude mcp list` and verify registration |
  155. | Expert not triggered | Try explicit: "Ask GPT to review..." or "Ask Gemini to review..." |
  156. ---
  157. ## Development
  158. ```bash
  159. git clone https://github.com/jarrodwatts/claude-delegator
  160. cd claude-delegator
  161. # Test locally without reinstalling
  162. claude --plugin-dir /path/to/claude-delegator
  163. ```
  164. See [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines.
  165. ---
  166. ## Acknowledgments
  167. Expert prompts adapted from [oh-my-opencode](https://github.com/code-yeongyu/oh-my-opencode) by [@code-yeongyu](https://github.com/code-yeongyu).
  168. ---
  169. ## License
  170. MIT — see [LICENSE](LICENSE)
  171. ---
  172. ## Star History
  173. [![Star History Chart](https://api.star-history.com/svg?repos=jarrodwatts/claude-delegator&type=Date&v=2)](https://star-history.com/#jarrodwatts/claude-delegator&Date)

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.