메뉴
Overview of Second Opinion commands
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Flow: Auto - Full Issue Lifecycle in One Shot - Complete end-to-end workflow: start worktree → analyze issue → ELI5 plan + necessity gate → implement ...
Run quality checks (lint + test + typecheck + security) without committing
Flow: Finish - Quality Gates, Commit, Push, and Create PR - Run quality checks, commit changes, push the branch, and create a pull request.
Flow Commands - Streamlined worktree-based development workflow. No locks, no Redis - just git.
Flow: Merge PR and Clean Up - Merge the current branch's PR, then clean up the worktree and branch.
Diagnose flow workflow setup and environment
| name | second-opinion-help |
| description | Overview of Second Opinion commands |
Generated from a Claude Code command. Where the procedure references these Claude-only surfaces, adapt as follows:
~/.codex/config.toml, or fall back to the referenced repo scripts and CLI entry points.Commands for AI-powered code review using multiple LLM models.
| Command | Description |
|---|---|
/second-opinion-start | Quick review with sensible defaults (file + model + depth) |
/second-opinion-models | Interactive model and depth selection with menus |
/second-opinion-help | This help overview |
# Interactive model selection
/second-opinion-models
# Review specific code
/second-opinion-models "review my auth middleware"
| Key | Model | Provider | Best For |
|---|---|---|---|
gemini-3-pro | Gemini 3.1 Pro | Comprehensive analysis | |
gemini-2.5-pro | Gemini 2.5 Pro | Stable, proven | |
claude-sonnet | Claude Sonnet 4.6 | Anthropic | Fast, excellent for code review |
claude-haiku | Claude Haiku 4.5 | Anthropic | Fastest Claude, cost-effective |
claude-opus | Claude Opus 4.6 | Anthropic | Most capable Claude |
codex | GPT-5.3 Codex | OpenAI | Default coding model |
codex-mini | GPT-5.2 Codex | OpenAI | Cost-effective coding |
o4-mini | o4-mini | OpenAI | Fast reasoning |
o3 | o3 | OpenAI | Advanced reasoning |
gpt-5.2 | GPT-5.2 | OpenAI | Latest GPT |
gpt-4o | GPT-4o | OpenAI | Fast multimodal |
You can also invoke the MCP tools directly without commands:
get_code_second_opinion - Single-model review (Gemini default)get_multi_model_second_opinion - Multi-model parallel reviewlist_available_models - Check which models are configuredcreate_session / consult - Multi-turn conversationscooneycw/mcp-second-opinion. It is opt-in and is not started or auto-registered by CPP.second-opinion MCP server registered in Claude Code, pointing at it. Installing via the plugin does NOT auto-register it, so register it yourself once the server is up:
claude mcp add second-opinion --transport http --url http://127.0.0.1:8080/mcp --scope user (use your Tailscale URL for a remote host)..mcp.json already registers second-opinion at project scope (http://127.0.0.1:8080/mcp for localhost, or a Tailscale URL).second-opinion ... Failed to connect or "1 error during load" on a fresh plugin install is expected until you opt in: the external server is not running and the plugin does not auto-register it. It is benign, the other commands are unaffected. Start the server and register it (below) to clear it.
Error -32602: Invalid request parameters usually means the server isn't running or second-opinion is not registered/pointing at it, not that parameters are wrong.
Fix: Make sure the external server is up and second-opinion is registered against it:
cooneycw/mcp-second-opinion checkout (see that repo's README).second-opinion as a streamable-http server at the right URL. Plugin install: claude mcp add second-opinion --transport http --url http://127.0.0.1:8080/mcp --scope user. CPP repo checkout: confirm the repo-root .mcp.json second-opinion entry targets the same URL (http://127.0.0.1:8080/mcp for localhost, or your Tailscale URL for a remote host).