GitHub Copilot CLI SKILL - use a variety of models (Gemini, GPT / Codex) from within Claude

0.README.md

GitHub Copilot CLI SKILL

Use models like Gemini 3 Pro, GPT-5.1, and GPT-5.1-Codex from within Claude by invoking GitHub Coplit CLI.

Installation

1. Create ~/.claude/skills/github-copilot
2. Save SKILL.md to ~/.claude/skills/github-copilot/SKILL.md

Usage

> Use Gemini to review the codebase and suggest a refactoring plan.

SKILL.md

name

description

github-copilot

Consult other AI models via GitHub Copilot CLI for second opinions, thorough analysis, or alternative perspectives. Supports Gemini 3 Pro Preview (gemini), GPT-5.1 (gpt), GPT-5.1-Codex (codex), and other non-Anthropic models. Use when user explicitly requests, when needing detailed analysis, when requiring additional help with an especially complex task, or when seeking alternative model perspectives.

GitHub Copilot CLI Integration

Invoke other AI models via GitHub Copilot CLI to obtain alternative perspectives and analysis. This skill acts as a transparent conduit: it passes context and prompts to the specified model via copilot CLI, then inserts the verbatim response into the conversation.

Core Function

This skill does NOT perform analysis itself. It:

1. Formats relevant context and the user's request into a prompt
2. Invokes copilot CLI with the specified model
3. Captures the complete response
4. Returns the response verbatim for integration into the conversation

Available Models

• gemini-3-pro-preview (default) - Shortcuts: "gemini"
• gpt-5.1 - Shortcuts: "gpt"
• gpt-5.1-codex - Shortcuts: "codex"

Default to gemini-3-pro-preview unless user specifies otherwise.

Note: Copilot may take longer to respond. Use appropriate timeout values (up to 23 minutes is acceptable).

Model Selection

• Default: Always use gemini-3-pro-preview
• User specifies a model: Map shortcuts to full model names (gemini → gemini-3-pro-preview, gpt → gpt-5.1, codex → gpt-5.1-codex)

Command Invocation

Execute copilot in the current working directory using this pattern:

copilot --model MODEL_NAME --allow-all-paths --allow-all-tools --log-level none -p "PROMPT_TEXT" 2>/dev/null

Required flags:

• --allow-all-paths - Allow access to all paths
• --allow-all-tools - Allow all tool usage
• --log-level none - Suppress log output

The command runs in the current working directory, providing copilot with the same context as Claude.

Prompt Construction

Construct prompts that provide:

1. Sufficient context - Include relevant conversation history, code snippets, or file contents
2. Clear request - State what analysis or output is needed
3. Specific details - Include constraints, requirements, or preferences

Keep prompts focused and relevant. Include only necessary context.

Response Handling

1. Execute the copilot command
2. Capture the complete output
3. Return the response in EXACTLY this format with no additions or modifications:

---
copilot-model: $MODEL
prompt: $PROMPT
---

$RESPONSE

Where:

• $MODEL = the actual model name used (e.g., "gemini-3-pro-preview")
• $PROMPT = the exact prompt sent to copilot
• $RESPONSE = the complete verbatim output from copilot

CRITICAL: Do not add any text before or after this code block. Do not editorialize, summarise, or modify the response in any way.

Example Usage Patterns

User requests alternative perspective

User: "Can you get a second opinion on this approach?" → Invoke copilot with gemini-3-pro-preview, including relevant context

User requests specific model

User: "What does GPT-5.1 think about this code?" → Invoke copilot with gpt-5.1, including the code and question

Thorough analysis needed

Task requires detailed investigation → Consider invoking gemini-3-pro-preview for comprehensive analysis

Output Format Example

---
copilot-model: gemini-3-pro-preview
prompt: Explain shallow vs deep copy in Python
---

A shallow copy creates a new object but references the same nested objects...
[complete response]

Important Notes

• This skill is a conduit, not an analyst
• Do not interpret, summarise, or modify responses
• Include sufficient context in prompts for meaningful responses
• The response quality depends entirely on the prompt quality
• Anthropic models (Claude) are available natively - do not use this skill for them