/consult-gemini

name: multi-ai-consultant:consult-gemini
description: Consult Google Gemini 3 Pro for second opinion with web research, extended thinking, and grounding. Use when need latest docs, architectural decisions, or security verification.
allowed-tools:
  - Bash
  - Read

Consult Gemini 3 Pro

You are being asked to consult Google Gemini 3 Pro for a second opinion on a coding problem or architectural decision.

Your Task

1. **Pre-flight Check**: Verify Gemini CLI is working 2. **Gather Context**: Determine what context to send (bug vs architecture) 3. **Build Prompt**: Create a clear, specific question with context 4. **Execute Consultation**: Call Gemini with locked configuration 5. **Parse Response**: Extract the analysis from JSON output 6. **Synthesize**: Compare Gemini's analysis with your own reasoning 7. **Present**: Show 5-part synthesis to user 8. **Log Cost**: Track tokens and cost

---

Step 1: Pre-flight Check

# Test if Gemini CLI is working
if ! gemini -p "test" &>/dev/null 2>&1; then
  echo "❌ Gemini CLI not working"
  echo ""
  echo "Possible issues:"
  echo "1. Gemini CLI not installed: npm install -g @google/generative-ai-cli"
  echo "2. API key not set: export GEMINI_API_KEY='your-key-here'"
  echo "3. API key invalid: Check https://aistudio.google.com/apikey"
  echo ""
  echo "After fixing, try the consultation again."
  exit 1
fi

If pre-flight check fails, stop here and show the error message to the user.

---

Step 2: Smart Context Selection

**For bug/error problems**:

• Include the specific file with the bug
• Include related files (imports, dependencies)
• Include error messages and stack traces
• Use `@path/to/file.ts` syntax

**For architecture decisions**:

• Include relevant source directories
• Include documentation if available
• Use `@src/` or `@path/to/feature/` syntax

**For general questions**:

• Use `@.` for entire codebase (if small)
• Or use `@src/` for source only

**Context size guidance**:

• Small bug: 1-3 files
• Architecture: 1-2 directories
• Full codebase: Only if <10k tokens estimated

---

Step 3: Build Prompt

The prompt must include: 1. **Problem statement**: What's wrong or what decision is needed 2. **Question**: Specific question for Gemini 3. **Your analysis**: What you've tried or considered 4. **Context**: File paths using `@<path>` syntax

**Format**:

Problem: [Brief description]
Question: [Specific question]

I've tried: [Your attempts]
Current status: [What's happening now]

Context: @path/to/file1.ts @path/to/file2.ts

---

Step 4: Execute Consultation

**Locked Configuration** (always use these flags):

gemini -m gemini-3-pro \
  --thinking \
  --google-search \
  --grounding \
  --output-format json

**Full command pattern**:

PROMPT=$(cat <<'EOF'
[Your formatted prompt here]
EOF
)

RESPONSE=$(echo "$PROMPT" | gemini \
  -m gemini-3-pro \
  --thinking \
  --google-search \
  --grounding \
  --output-format json 2>/dev/null)

EXIT_CODE=$?
if [ $EXIT_CODE -ne 0 ]; then
  echo "❌ Gemini API call failed (exit code: $EXIT_CODE)"
  echo "Check your API key and internet connection"
  exit 1
fi

---

Step 5: Parse Response

# Extract the analysis text
ANALYSIS=$(echo "$RESPONSE" | jq -r '.response.text')

# Extract token counts for cost tracking
INPUT_TOKENS=$(echo "$RESPONSE" | jq -r '.stats.inputTokens')
OUTPUT_TOKENS=$(echo "$RESPONSE" | jq -r '.stats.outputTokens')

# Check if parsing succeeded
if [ -z "$ANALYSIS" ] || [ "$ANALYSIS" == "null" ]; then
  echo "❌ Failed to parse Gemini response"
  echo "Raw response:"
  echo "$RESPONSE"
  exit 1
fi

---

Step 6: Synthesize (5-Part Format)

You MUST compare Gemini's analysis with your own reasoning. DO NOT just parrot Gemini's response.

**Required sections**:

🤖 My Analysis

• Your original reasoning
• What you tried
• What you think the issue is
• Your proposed solution

💎 Gemini's Analysis

• Gemini's complete response
• Include their reasoning process
• Include any web research findings
• Include any grounding citations

🔍 Key Differences

• **Agreement**: Where you and Gemini align
• **Divergence**: Where you disagree (explain why)
• **Gemini's catch**: Things Gemini found that you missed
• **Your edge**: Things you considered that Gemini didn't
• **Web research value**: What Google Search added

⚡ Synthesis

• Combine both perspectives
• Identify root cause (if bug)
• Weigh trade-offs (if architecture)
• Note any conflicting advice and why

✅ Recommended Action

• Specific, actionable next steps
• File paths and line numbers
• Expected outcome
• Risk assessment

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

---

Step 7: Log Cost

# Calculate cost (Gemini 3 Pro pricing)
# Input: $0.000015/token, Output: $0.00006/token (example rates, verify current)
COST=$(echo "scale=4; ($INPUT_TOKENS * 0.000015) + ($OUTPUT_TOKENS * 0.00006)" | bc)

# Create log directory if needed
mkdir -p ~/.claude/ai-consultations

# Log consultation
echo "$(date -Iseconds),gemini,gemini-3-pro,$INPUT_TOKENS,$OUTPUT_TOKENS,$COST,$(pwd)" \
  >> ~/.claude/ai-consultations/consultations.log

# Show cost to user
echo ""
echo "---"
echo "Cost: \$$COST ($INPUT_TOKENS input + $OUTPUT_TOKENS output tokens)"
echo "Consultation logged: ~/.claude/ai-consultations/consultations.log"

---

Privacy & Security

**Always respect**:

• `.gitignore` (automatically excluded by Gemini CLI)
• `.geminiignore` (if present in project root)

**Never send**:

• `.env*` files
• `*secret*` files
• `*credentials*` files
• `node_modules/`
• Build artifacts

**If user has sensitive files in context**: Warn before sending and ask permission.

---

Error Handling

**Common errors**: 1. **API key invalid**: Show how to set `GEMINI_API_KEY` 2. **Context too large**: Suggest narrowing context 3. **JSON parsing fails**: Fall back to plain text output 4. **Rate limit**: Suggest waiting 60 seconds 5. **Network error**: Check in