Gemini

Gemini Skill Guide

When to Use Gemini

• Frontend Development: Building UI components and pages (Gemini 3 Pro produces exceptional frontend quality). When the frontend-design skill is installed, load it and pass its guidelines to Gemini for design-driven results.
• Code Review: Comprehensive reviews across multiple files
• Plan Review: Analysing architectural plans, technical specifications, or project roadmaps
• Big Context Processing: Tasks requiring >200k tokens of context (entire codebases, documentation sets)
• Multi-file Analysis: Understanding relationships and patterns across many files

Running a Task

1. Use the latest Gemini 3 Pro model by default (currently gemini-3-pro-preview at time of writing). Ask the user which model to use only if they want to change from the default. See the Model Selection table below for alternatives.

2. Select the approval mode based on execution context:

   • yolo -- Required for background/automated tasks (Claude Code tool calls, CI/CD). Prevents hung processes.
   • auto_edit -- Auto-approves edit tools only. Good for code reviews with suggestions.
   • default -- Prompts for approval. Only works in interactive terminal sessions. Never use in background shells.

3. Assemble the command:

   • -m, --model <MODEL> -- Model selection
   • --approval-mode <default|auto_edit|yolo> -- Control tool approval
   • -y, --yolo -- Shorthand for --approval-mode yolo
   • -i, --prompt-interactive "prompt" -- Execute prompt then continue interactively
   • --include-directories <DIR> -- Additional directories to include in workspace
   • -s, --sandbox -- Run in sandbox mode for isolation

4. For background/automated tasks, always use --approval-mode yolo and wrap with timeout:

   timeout 300 gemini -m gemini-3-pro-preview --approval-mode yolo "Your prompt here"
   

5. Run the command and capture the output.

6. After Gemini completes, inform the user: "The Gemini analysis is complete. You can start a new Gemini session for follow-up analysis."

Frontend Design Integration

Gemini 3 Pro is exceptionally strong at frontend design. When the task involves building UI components, pages, or visual interfaces:

1. Check whether the frontend-design skill is installed.
2. If installed, load it and include its design guidelines (typography, colour, motion, accessibility) in the prompt sent to Gemini. This gives Gemini concrete aesthetic direction and prevents generic "AI slop" output.
3. If not installed, proceed with Gemini alone -- it still produces high-quality frontend code, but without the opinionated design system the frontend-design skill provides.

The combination works well because Gemini 3 Pro handles complex UI reasoning and the frontend-design skill supplies the design constraints that turn competent code into distinctive interfaces.

Task Checklist

- [ ] 1. Confirm model (default: gemini-3-pro-preview)
- [ ] 2. Select approval mode (default: yolo for background)
- [ ] 3. Assemble command with flags
- [ ] 4. Wrap with timeout for background tasks
- [ ] 5. Run command
- [ ] 6. Summarise outcome to user
- [ ] 7. Ask user about next steps

Quick Reference

Model Selection

Default to the latest Gemini 3 Pro model (currently gemini-3-pro-preview). Only switch when the user explicitly asks for speed, lower cost, or a legacy model.

All models support 1M input tokens and 64-65k output tokens.

Examples

Claude assembles:

timeout 300 gemini -m gemini-3-pro-preview --approval-mode yolo \
  "Perform a comprehensive code review focusing on:
   1. Security vulnerabilities
   2. Performance issues
   3. Code quality and maintainability
   4. Best practices violations"

After completion: "Gemini found 3 potential issues: an SQL injection risk in db/queries.ts, an unvalidated redirect in auth/callback.ts, and a missing rate limiter on the /api/upload endpoint. You can start a new Gemini session for deeper analysis of any finding."

Claude assembles:

timeout 300 gemini -m gemini-3-pro-preview --approval-mode yolo \
  "Analyse the entire codebase to understand:
   1. Overall architecture
   2. Key patterns and conventions
   3. Potential technical debt
   4. Refactoring opportunities"

After completion: Summarises the architectural findings and asks the user about next steps.

Claude checks: frontend-design skill is installed. Loads it and extracts design guidelines.

Claude assembles:

timeout 300 gemini -m gemini-3-pro-preview --approval-mode yolo \
  "Build a responsive dashboard component with:
   1. Clean, modern UI design
   2. Smooth animations and transitions
   3. Accessible markup (ARIA labels, keyboard navigation)
   4. Mobile-first responsive layout

   Design constraints (from frontend-design skill):
   - Use distinctive typography (avoid Inter/Roboto/Arial)
   - Define colours as CSS custom properties, no hardcoded hex
   - Add entry animations and hover/scroll interactions
   - Respect prefers-reduced-motion
   - Every element must earn its place"

After completion: Reviews generated code with git diff, verifies accessibility and design quality, then reports results.

Claude assembles (uses Flash for speed):

timeout 120 gemini -m gemini-3-flash --approval-mode yolo \
  "Review this file for bugs and anti-patterns"

Following Up

• Gemini CLI sessions are one-shot or interactive. There is no built-in resume.
• For follow-up analysis, start a new Gemini session with context from previous findings.
• When proposing follow-up actions, restate the chosen model and approval mode.

Error Handling

• Stop and report failures whenever gemini --version or a Gemini command exits non-zero. Request direction before retrying.
• Before using --approval-mode yolo for the first time in a session, confirm with the user.
• When output includes warnings or partial results, summarise them and ask how to adjust.
• For hung process detection and resolution, see references/troubleshooting.md.

Tips for Large Context Processing

1. Be specific: Provide clear, structured prompts for what to analyse.
2. Use include-directories: Explicitly specify all relevant directories.
3. Break down complex tasks: Even with 1M tokens, structured analysis is more effective.
4. Save findings: Ask Gemini to output structured reports that can be saved for reference.

CLI Version

Requires a Gemini CLI version that supports the Gemini 3 model family (v0.16.0 was the minimum at time of writing). Check with gemini --version and upgrade if the command fails with an unknown-model error.