Skip to content

Gemini CLI Cheatsheet

Quick reference for everything covered in this workshop.

Last updated: 2026-05-05 · Source verified against gemini-cli repository

Installation

npm install -g @google/gemini-cli
gemini                     # Launch interactive mode
gemini --version           # Check version

Keyboard Shortcuts

Shortcut Action
Tab Accept suggested edit
Shift+Tab Cycle through options
Ctrl+G External editor (edit prompt or plan)
Ctrl+C Cancel current operation
↑ / ↓ Navigate prompt history

Slash Commands

Command Description
/plan Toggle Plan Mode (read-only research)
/stats Show token usage and model info
/clear Clear context and start fresh
/tools List available tools
/resume Resume a previous session
/rewind Roll back to a previous state
/restore Restore from a checkpoint (requires checkpointing enabled)
/memory show Show saved memories
/memory reload Reload memory from source files
/memory list List all GEMINI.md files in use
/memory inbox Review auto-extracted memory candidates (requires experimental.autoMemory: true)
~~/memory add "..."~~ ~~Add a memory~~ — removed in v0.41.1, use natural language instead: "Remember that..." (details)
/hooks panel Show hook execution status
/skills list List available skills
/extensions list List installed extensions
/commands List custom commands

Headless Mode

# Simple prompt
gemini -p "Explain this code"

# Structured output
gemini -p "List endpoints as JSON" --output-format json

# Pipe input
cat error.log | gemini -p "Diagnose this error"

# Pipe code
cat file.js | gemini -p "Review this code for bugs"

GEMINI.md Hierarchy

~/.gemini/GEMINI.md          # Global preferences
./GEMINI.md                  # Project conventions
./backend/GEMINI.md          # Subdirectory rules
./frontend/GEMINI.md         # Subdirectory rules

Import syntax

@./docs/coding-standards.md
@./docs/architecture.md

See GEMINI.md reference for full syntax.

Subagents

# Built-in
@codebase_investigator Map the call chain for the login endpoint

# Custom (defined in .gemini/agents/)
@security-scanner Review auth middleware for vulnerabilities

Subagent definition (.gemini/agents/my-agent.md)

---
model: gemini-3.1-flash-lite-preview
tools:
  - read_file
  - list_directory
---
You are a specialist in...

Conductor Extension

# Install
gemini extensions install https://github.com/gemini-cli-extensions/conductor

# Set up project context
/conductor:setup prompt="Project description..."

# Create a feature track
/conductor:newTrack prompt="Feature description..."

# Implement the current track
/conductor:implement

Policy Engine (TOML)

# Deny reading secrets
[[rule]]
toolName = "read_file"
argsPattern = '"file_path":".*\.env"'
decision = "deny"
priority = 100
deny_message = "Reading .env files is not allowed."

# Allow running tests
[[rule]]
toolName = "run_shell_command"
commandPrefix = "npm test"
decision = "allow"
priority = 50

# Default: ask human
[[rule]]
toolName = "*"
decision = "ask_user"
priority = 1

See Policy Engine reference for the full schema and Secure Gemini CLI with the Policy Engine for a practical walkthrough.

Hooks

Settings.json hook config

{
  "hooks": {
    "BeforeTool": [{
      "matcher": "write_file|replace",
      "hooks": [{
        "name": "my-hook",
        "type": "command",
        "command": "$GEMINI_PROJECT_DIR/.gemini/hooks/my-hook.sh",
        "timeout": 3000
      }]
    }]
  }
}

Hook script template

#!/usr/bin/env bash
input=$(cat)
filepath=$(echo "$input" | jq -r '.tool_input.file_path // ""')

# Allow (default)
echo '{}'

# Deny with reason
echo '{"decision":"deny","reason":"Blocked because..."}'

# Inject context
echo '{"systemMessage":"Remember to..."}'

Hook events

SessionStart → BeforeAgent → BeforeModel → BeforeToolSelection →
AfterModel → BeforeTool → AfterTool → AfterAgent → PreCompress →
Notification → SessionEnd

See Hooks reference for the complete event lifecycle.

MCP Servers

{
  "mcpServers": {
    "github": {
      "command": "npx",
      "args": ["-y", "github-mcp-server"],
      "env": {
        "GITHUB_PERSONAL_ACCESS_TOKEN": "$GITHUB_TOKEN"
      }
    }
  }
}

GitHub Actions

- uses: google-github-actions/run-gemini-cli@v1
  with:
    prompt: "Review this PR for code quality and security"

Auth Options

# Personal (free tier)
gemini   # OAuth flow

# Vertex AI (enterprise)
gcloud auth application-default login
# + configure .gemini/settings.json with auth.provider = "vertex-ai"

Useful Patterns

# Smart commit
git diff --cached | gemini -p "Generate a conventional commit message"

# Code review
git diff main...HEAD | gemini -p "Review these changes"

# Generate docs
gemini -p "Generate JSDoc for all exports in backend/controllers/"

# Batch processing
for f in src/*.js; do gemini -p "Add TypeScript types" < "$f"; done

Extensions

# Install from GitHub
gemini extensions install https://github.com/owner/repo

# Install a specific version
gemini extensions install https://github.com/owner/repo --ref v1.2.0

# List installed extensions
gemini extensions list
/extensions list   # from interactive mode

# Update all extensions
gemini extensions update --all

# Create from a template
gemini extensions new my-extension mcp-server

# Develop locally (symlink — changes reflected immediately)
gemini extensions link .

# Disable for this workspace only
gemini extensions disable my-extension --scope workspace

Notable Community Extensions

# Conductor (spec-driven development) — already in UC1
gemini extensions install https://github.com/gemini-cli-extensions/conductor

# Superpowers (TDD, code review, subagent-driven development)
gemini extensions install https://github.com/obra/superpowers

# Oh-My-Gemini-CLI (multi-agent orchestration framework)
gemini extensions install https://github.com/Joonghyun-Lee-Frieren/oh-my-gemini-cli

# Google Workspace CLI (optional — requires Workspace auth)
gemini extensions install https://github.com/googleworkspace/cli

Browse community extensions: geminicli.com/extensions/browse

Publish your own: Add gemini-cli-extension topic to your GitHub repo + tag a release.