multi-ai-consultant

Multi-AI Consultant

Consult external AIs for second opinions when Claude Code is stuck or making critical decisions.

What This Skill Does

This skill enables future Claude Code sessions to consult other AIs when:

• Stuck on a bug after one failed attempt
• Making architectural decisions
• Security concerns need validation
• Fresh perspective needed

Key innovation: Uses existing CLI tools (gemini, codex) instead of building MCP servers - much simpler and more maintainable.

When to Use This Skill

Automatic Triggers (No User Action Needed)

Claude Code should automatically suggest using this skill when:

1. After 1 failed debugging attempt
2. Tried one approach to fix a bug
3. Still not working or different error
4. → Suggest: "Should I consult [Gemini|Fresh Claude] for a second opinion?"

1. Before architectural decisions
2. Significant design choices (state management, routing, data flow)
3. Framework selection
4. Database schema design
5. → Auto-consult (mention to user): "Consulting Gemini for architectural validation..."

1. Security changes
2. Authentication logic
3. Authorization rules
4. Cryptography
5. Input validation
6. → Auto-consult: "Consulting Gemini to verify security approach..."

1. When uncertain
2. Multiple valid approaches
3. Trade-offs not clear
4. Conflicting advice in documentation
5. → Suggest: "Would you like me to consult another AI for additional perspective?"

Manual Invocation (User Commands)

User can explicitly request consultation with:

• /consult-gemini [question] - Gemini 2.5 Pro with thinking, search, grounding
• /consult-codex [question] - OpenAI GPT-4 via Codex CLI (repo-aware)
• /consult-claude [question] - Fresh Claude subagent (free, fast)
• /consult-ai [question] - Router that asks which AI to use

The Three AIs

| AI | Tool | When to Use | Special Features | Cost | |----|------|-------------|------------------|------| | Gemini 2.5 Pro | gemini CLI | Web research, latest docs, thinking | Google Search, extended reasoning, grounding | ~$0.10-0.50 | | OpenAI GPT-4 | codex CLI | Repo-aware analysis, code review | Auto-scans directory, OpenAI reasoning | ~$0.05-0.30 | | Fresh Claude | Task tool | Quick second opinion, budget-friendly | Same capabilities, fresh perspective | Free |

For detailed AI comparison: Load references/ai-strengths.md when choosing which AI to consult for specific use cases.

How It Works

Architecture

Claude Code encounters bug/decision
        ↓
Suggests consultation (or user requests)
        ↓
User approves
        ↓
Execute appropriate slash command
        ↓
CLI command calls external AI
        ↓
Parse response
        ↓
Synthesize: Claude's analysis + External AI's analysis
        ↓
Present 5-part comparison
        ↓
Ask permission to implement

The 5-Part Synthesis Format

Every consultation must follow this format (prevents parroting):

1. 🤖 My Analysis - Claude's original reasoning and attempts
2. 💎/🔷/🔄 Other AI's Analysis - External AI's complete response
3. 🔍 Key Differences - Agreement, divergence, what each AI caught/missed
4. ⚡ Synthesis - Combined perspective, root cause, trade-offs
5. ✅ Recommended Action - Specific next steps with file paths/line numbers

End with: "Should I proceed with this approach?"

Setup

For complete installation guide: Load references/setup-guide.md when installing CLIs, configuring API keys, or setting up templates.

Quick setup:

1. Install CLIs: bun add -g @google/generative-ai-cli (Gemini), bun add -g codex (Codex, optional)
2. Set API keys: export GEMINI_API_KEY="...", export OPENAI_API_KEY="..."
3. Install skill: Symlink to ~/.claude/skills/multi-ai-consultant
4. Copy templates: GEMINI.md, codex.md, .geminiignore to project root
5. Verify: gemini -p "test", codex exec - --yolo

Get API keys:

• Gemini: https://aistudio.google.com/apikey
• OpenAI: https://platform.openai.com/api-keys

Usage

For detailed examples: Load references/usage-examples.md when learning consultation workflows or seeing real-world scenarios.

Quick examples:

• Bug: After 1 failed attempt → /consult-gemini for web-researched solution
• Architecture: Design decision → /consult-gemini for latest best practices
• Code review: Refactoring validation → /consult-codex for repo-aware consistency check
• Quick opinion: Sanity check → /consult-claude for free fresh perspective

5 detailed examples available:

1. JWT authentication bug (saved ~30 min, found platform-specific issue)
2. State management choice (informed decision with 2025 patterns)
3. Refactoring review (found 3 consistency issues)
4. Security validation (found 2 critical issues via OWASP 2025)
5. Multi-AI workflow (high-stakes database choice)

Slash Commands

For complete command reference: Load references/commands-reference.md when needing detailed syntax, options, or cost tracking information.

Quick command overview:

/consult-gemini [question]

• Use: Web research, latest docs, extended thinking
• Features: Google Search, grounding, thinking mode
• Cost: ~$0.10-0.50
• Example: /consult-gemini Is this JWT secure by 2025 standards?

/consult-codex [question]

• Use: Repo-aware analysis, code review
• Features: Auto-scans directory, consistency checks
• Cost: ~$0.05-0.30
• Example: /consult-codex Review for performance bottlenecks

/consult-claude [question]

• Use: Quick second opinion, budget-friendly
• Features: Free, fast, fresh perspective
• Cost: Free
• Example: /consult-claude Am I missing something obvious?

/consult-ai [question]

• Use: Router (recommends which AI to use)
• Features: Analyzes question, suggests best AI
• Cost: Varies by chosen AI
• Example: /consult-ai How should we structure this architecture?

Templates

Templates customize AI behavior for consultations (auto-loaded from project root):

1. GEMINI.md - System instructions for Gemini (enforces 5-part format, web search)
2. codex.md - System instructions for Codex (enforces repo-aware analysis)
3. .geminiignore - Privacy exclusions beyond .gitignore
4. consultation-log-parser.sh - View consultation history (optional)

Installation: Copy from ~/.claude/skills/multi-ai-consultant/templates/ to project root

Privacy & Security

Automatic protection:

• Both CLIs respect .gitignore automatically
• Create .geminiignore for extra exclusions (.env*, secret, credentials)
• Pre-consultation check warns if sensitive patterns detected

Privacy best practices:

• Always configure .gitignore properly
• Create .geminiignore for extra safety
• Use smart context selection (specific files, not entire repo)
• Verify what will be sent: git status --ignored

For detailed privacy configuration: Load references/setup-guide.md when setting up .geminiignore or privacy exclusions.

Cost Tracking

Every consultation logged to ~/.claude/ai-consultations/consultations.log

Log format: timestamp,ai,model,input_tokens,output_tokens,cost,project_path

View logs: consultation-log-parser.sh --summary

Example output:

Total consultations: 47
Gemini: 23 ($4.25), Codex: 12 ($1.85), Fresh Claude: 12 ($0.00)
Total cost: $6.10

For detailed cost tracking: Load references/commands-reference.md when viewing logs, calculating costs, or managing budgets.

Common Issues

For complete troubleshooting: Load references/troubleshooting.md when encountering errors or setup issues.

Top 5 issues:

1. CLI not installed: gemini: command not found → Fix: bun add -g @google/generative-ai-cli
2. API key invalid: Authentication failed → Fix: `export GEMINI_API_KEY=".