---

name: second-opinion description: Get second opinions from external AI agents (OpenAI, Gemini, GitHub). Use when user says "second opinion", "ask other AIs", "what would GPT say", "check with others", or wants to validate an approach.

Second Opinion

Query external agents, get answer, act on it.

Commands

CRITICAL: Execution Flow

Step 1: Check for Installed Agent Plugins

FIRST, check if any agent MCP tools are available. Look for these tools:

Note: mcp__anthropic__agent_query is excluded when running in Claude Code.

If NONE of these tools exist, display this message and STOP immediately:

CopyNo second-opinion agents available.

Run `/second-opinion setup` to configure authentication:
- GitHub: Uses OAuth via `gh auth login`
- OpenAI: Enter API key interactively
- Gemini: Enter API key interactively

Do NOT proceed further if no agents are available.

Step 2: Gather Opinions (use opinion-gatherer agent)

If agents ARE installed, spawn the opinion-gatherer agent with:

• The user's query/question
• Your own proposal/answer to the question
• Any relevant context

The agent will:

1. Output "Gathering second opinions..."
2. Query each available agent sequentially, showing progress:
   • " Asking OpenAI..."
   • " Asking Gemini..."
   • " Asking GitHub..."
3. Return aggregated results (including actual model names from responses)

Step 3: Analyze and Decide

After receiving responses from the opinion-gatherer agent, display:

CopyAnalyzing responses against my proposal...

Then apply this decision logic:

Output Guidelines

Do NOT:

• Attribute opinions to specific agents by name
• Show verbose breakdowns of each response
• Summarize what each agent said individually

DO:

• Return actionable next step
• Keep it brief
• Only ask user when truly stuck

Output Examples

Consensus:

CopyUse dependency injection. Add the service as a constructor param.

Decided:

CopyGoing with extracted module approach.

_Cleaner separation, easier to test._

Needs input (rare):

CopyNeed your input on this one:

- **Sync**: Simpler but blocks UI
- **Async**: More complex but responsive

First-Time Setup

Run /second-opinion setup to configure agents interactively.