From ai-panel
Consults other AI models (Gemini, GPT) for second opinions on architecture, approach, and trade-offs. Use during planning or when a decision benefits from multiple perspectives.
How this skill is triggered — by the user, by Claude, or both
Slash command
/ai-panel:ai-panelThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
- **Plan mode**: Get alternative perspectives on architecture, technology choices, or trade-offs before committing to an approach.
When NOT to use: Trivial decisions, obvious fixes, or when you're already highly confident. Don't waste premium requests on things that don't benefit from diversity of thought.
| Model | Command | Strengths |
|---|---|---|
| Google Gemini | gemini -p "prompt" | Web-grounded knowledge, long context, research questions |
| GPT (via Copilot) | copilot -p "prompt" --model gpt-5.2 | Strong reasoning, code generation, structured output |
Important: Copilot defaults to Claude Sonnet — always pass --model gpt-5.2 to get a genuinely different perspective. The point of the panel is model diversity.
Always call both models in parallel (two Bash calls in a single message) and wrap in timeout to prevent hangs.
# Gemini (30s timeout)
timeout 30 gemini -p "Context: ... Question: ... Be concise, 3-5 key points."
# GPT via Copilot (30s timeout)
timeout 30 copilot -p "Context: ... Question: ... Be concise, 3-5 key points." --model gpt-5.2 --allow-all-tools
/second-opinion)Architecture decisions, technology choices, approach comparison — anything where genuine disagreement is valuable.
Strip the Loaded cached credentials. prefix line if present. The useful response starts after it.
Strip the trailing usage stats block. Everything from the blank line before Total usage est: onwards should be removed. Only keep the actual response content.
If a model times out (30s), note it in the output and proceed with whichever model responded. Don't retry — the user can re-run if needed.
Format the panel's output clearly:
### Gemini
<cleaned response>
### GPT
<cleaned response>
### Synthesis
- **Agreement**: Where both models align
- **Disagreement**: Where they differ and why
- **Recommendation**: Your own assessment, weighing all perspectives
Always give your own recommendation last. The panel informs your judgement — it doesn't replace it.
npx claudepluginhub sling86/ai-panel --plugin ai-panelQueries AI models via OpenRouter, Gemini, or OpenAI APIs for second opinions on code, architecture, strategy, or prompting. Supports consensus, single opinion, and devil's advocate modes.
Consults Google's Gemini model for second opinions on coding tasks like debugging tricky issues, evaluating architecture trade-offs, reviewing code, and brainstorming alternatives.
Consults external LLMs (OpenAI Codex, Google Gemini) via CLIs for second opinions on architecture, design decisions, model selection, and approach comparisons.