Second Opinion
Query external agents, get answer, act on it.
Commands
| Command | Description |
|---|---|
/second-opinion | Get a second opinion on current context |
/second-opinion setup | Configure which agents to enable |
/second-opinion status | Show current configuration |
Behavior
| Situation | Action |
|---|---|
| Consensus | Proceed silently |
| No consensus, decidable | Decide best option, proceed |
| No consensus, critical | Ask user (rare) |
Do NOT:
- •Attribute opinions to specific agents
- •Show verbose breakdowns
- •Summarize what each agent said
DO:
- •Return actionable next step
- •Keep it brief
- •Only ask user when truly stuck
Usage
json
{
"query": "Is this the right approach for handling auth?",
"context": "<code or description>"
}
Output Examples
Consensus:
code
Use dependency injection. Add the service as a constructor param.
Decided:
code
Going with extracted module approach. _Cleaner separation, easier to test._
Needs input (rare):
code
Need your input on this one: - **Sync**: Simpler but blocks UI - **Async**: More complex but responsive
First-Time Setup
If no agents are configured, run /second-opinion setup to:
- •Select which AI agents to enable (interactive)
- •Paste API keys when prompted (stored automatically)
- •GitHub token auto-detected from
gh auth tokenif installed
No file editing required - everything is configured interactively.