Copilot CLI vs Gemini CLI Tool Names for OpenAI Plugins: Complete Mapping Guide

Copilot CLI uses view, create, and edit while Gemini CLI uses read_file, write_file, and replace for the same generic plugin actions defined in the OpenAI Plugins repository.

The OpenAI Plugins repository defines skills using generic Claude Code tool names like Read, Write, and Bash. When these plugins execute on specific platforms, the Superpowers framework translates these generic commands into platform-specific CLI tool names automatically using data-driven mapping tables.

Generic Tool Names in OpenAI Plugins

Plugin skills in the OpenAI Plugins repository are authored once using standardized Claude Code tool names. These generic abstractions include common operations like Read for file access, Write for file creation, Edit for modifications, and Bash for command execution.

According to the source code in plugins/superpowers/skills/using-superpowers/SKILL.md, these generic names serve as the universal interface for skill definitions. When a skill runs on a specific platform, the Superpowers framework inspects the active CLI environment and substitutes the generic tool calls with the appropriate platform-specific equivalents.

Copilot CLI Tool Name Mapping

The GitHub Copilot CLI mapping lives in plugins/superpowers/skills/using-superpowers/references/copilot-tools.md. Key translations include:

• Read → view (file reading)
• Write → create (file creation)
• Edit → edit (file editing)
• Bash → bash (command execution)
• Grep → grep (content search)
• Glob → glob (filename search)
• Task → task (sub-agent dispatch)
• TodoWrite → sql (SQLite operations on todos table)
• WebSearch → web_fetch with search-engine URLs

Copilot CLI also supports async shell sessions through additional tools like write_bash and read_bash, allowing long-running commands to execute without blocking the skill.

Gemini CLI Tool Name Mapping

The Gemini CLI mapping is maintained in plugins/superpowers/skills/using-superpowers/references/gemini-tools.md. The platform uses more verbose, descriptive tool names:

• Read → read_file
• Write → write_file
• Edit → replace
• Bash → run_shell_command
• Grep → grep_search
• Glob → glob
• Task → @agent-name syntax (e.g., @generalist)
• TodoWrite → write_todos
• WebSearch → google_web_search

Unlike Copilot CLI, Gemini CLI does not support async shell sessions—all commands run synchronously.

Key Differences Between Copilot CLI and Gemini CLI

Several functional gaps exist beyond simple naming conventions:

Sub-agent Dispatch: Copilot CLI uses the task tool with a skill identifier, while Gemini CLI requires the @agent-name syntax (e.g., @generalist) with the prompt passed as an argument.

Todo Management: Copilot CLI translates TodoWrite into raw SQLite statements against a todos table, whereas Gemini CLI provides the dedicated write_todos command.

Web Search: Copilot CLI has no native web search tool; the Superpowers framework falls back to web_fetch with search-engine URLs. Gemini CLI exposes google_web_search directly.

Async Operations: Only Copilot CLI supports asynchronous shell execution through write_bash and read_bash. Gemini CLI lacks async shell capabilities entirely.

How Tool Name Translation Works at Runtime

The translation mechanism follows a four-step architectural flow implemented in the Superpowers skill:

1. Platform Detection: The framework identifies whether the skill is running on Copilot CLI, Gemini CLI, or another supported platform.

2. Table Lookup: Based on the detected platform, the system loads either copilot-tools.md or gemini-tools.md from the references directory.

3. Name Substitution: Generic Claude Code tool names are replaced with platform-specific equivalents before command execution.

4. Execution: The translated command runs in the target CLI environment, with results returned to the skill's prompt context.

Because the mapping is data-driven, adding support for a new CLI only requires creating a new reference table; the core skill code in SKILL.md remains unchanged.

Code Examples: Generic Skills to Platform-Specific Commands

Reading a File

Generic skill request:

Read path/to/file.txt

Copilot CLI translation:

view path/to/file.txt

Gemini CLI translation:

read_file path/to/file.txt

Writing a File

Generic skill request:

Write path/to/file.txt "Content here"

Copilot CLI execution:

create path/to/file.txt "Content here"

Gemini CLI execution:

write_file path/to/file.txt "Content here"

Running Shell Commands

Generic skill request:

Bash echo "Hello, world!"

Copilot CLI (with async support):

bash async:true "echo 'Hello, world!'"

Gemini CLI (synchronous only):

run_shell_command "echo 'Hello, world!'"

Dispatching Sub-agents

Generic skill request:

Task superpowers:code-reviewer <<PROMPT>>

Copilot CLI translation:

task superpowers:code-reviewer <<PROMPT>>

Gemini CLI translation:

@generalist "You are a code reviewer. <<PROMPT>>"

Updating Task Tracking

Generic skill request:

TodoWrite "Fix typo in README"

Copilot CLI (SQLite approach):

sql "INSERT INTO todos (title) VALUES ('Fix typo in README')"

Gemini CLI (dedicated tool):

write_todos "Fix typo in README"

Summary

• Copilot CLI uses short commands like view, create, and edit, while Gemini CLI uses descriptive names like read_file, write_file, and replace.
• The OpenAI Plugins repository stores canonical mappings in copilot-tools.md and gemini-tools.md within the references directory.
• Skills are authored once using generic Claude Code tool names, then translated at runtime by the Superpowers skill.
• Copilot CLI supports async shell sessions (write_bash, read_bash) and SQLite-based todos, while Gemini CLI runs commands synchronously and uses dedicated tools like write_todos and google_web_search.

Frequently Asked Questions

Where are the Copilot CLI and Gemini CLI tool mappings stored?

The mappings live in the references directory of the Superpowers skill. Specifically, plugins/superpowers/skills/using-superpowers/references/copilot-tools.md contains the Copilot CLI translations, and plugins/superpowers/skills/using-superpowers/references/gemini-tools.md contains the Gemini CLI equivalents.

Why do the same plugin skills use different commands on different CLIs?

The OpenAI Plugins architecture uses generic Claude Code tool names (like Read and Write) as a universal abstraction. When skills execute, the Superpowers framework consults platform-specific lookup tables to translate these generic names into the actual CLI commands available on Copilot CLI or Gemini CLI, ensuring compatibility without rewriting skill logic.

How does async bash execution differ between Copilot CLI and Gemini CLI?

Copilot CLI supports asynchronous shell sessions through dedicated tools like write_bash and read_bash, allowing skills to execute long-running commands without blocking. Gemini CLI lacks async shell capabilities entirely—all commands run synchronously using run_shell_command.

Where is the translation logic implemented in the source code?

The runtime translation logic is implemented in plugins/superpowers/skills/using-superpowers/SKILL.md. This file handles platform detection, loads the appropriate reference table (either Copilot or Gemini), and performs the substitution of generic tool names to platform-specific CLI commands before execution.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}