feat(skills): Add gemini-tools collection
- gemini-consultant: Second opinion with web search + vision - gemini-second-opinion: Independent technical analysis - gemini-image-generation: AI image generation (nano-banana CLI) - Includes consult.py script - Complements existing gemini-consultation skill - Updated SKILLS-INDEX.md Chronicler #73
This commit is contained in:
Claude
@@ -283,6 +283,30 @@
|
||||
|
||||
---
|
||||
|
||||
### gemini-tools (collection)
|
||||
**Location:** `docs/skills/gemini-tools/`
|
||||
**Source:** skill.fish (melodic-software, stared, dnvriend)
|
||||
**Triggers:** Gemini, second opinion, consultant, image generation, web search, fact verification
|
||||
|
||||
**Purpose:** Google Gemini integration tools
|
||||
|
||||
**What It Covers:**
|
||||
- **gemini-consultant.md** — Second opinion with web search + vision (includes consult.py)
|
||||
- **gemini-second-opinion.md** — Independent technical analysis
|
||||
- **gemini-image-generation.md** — AI image generation with nano-banana CLI
|
||||
|
||||
**Prerequisites:**
|
||||
- Requires `GEMINI_API_KEY` environment variable
|
||||
- API key from https://aistudio.google.com/app/apikey
|
||||
|
||||
**Read This When:**
|
||||
- Need real-time web information
|
||||
- Validating architecture decisions
|
||||
- Cross-checking Claude's analysis
|
||||
- Generating marketing images/thumbnails
|
||||
|
||||
---
|
||||
|
||||
### minecraft-bukkit-pro
|
||||
**Location:** `docs/skills/minecraft-bukkit-pro/SKILL.md`
|
||||
**Source:** skill.fish (sickn33/antigravity-awesome-skills)
|
||||
@@ -393,6 +417,12 @@ docs/skills/
|
||||
│ └── SKILL.md
|
||||
├── gemini-consultation/
|
||||
│ └── SKILL.md
|
||||
├── gemini-tools/
|
||||
│ ├── README.md
|
||||
│ ├── consult.py
|
||||
│ ├── gemini-consultant.md
|
||||
│ ├── gemini-image-generation.md
|
||||
│ └── gemini-second-opinion.md
|
||||
├── minecraft-bukkit-pro/
|
||||
│ └── SKILL.md
|
||||
├── minecraft-mod-dev/
|
||||
|
||||
123
docs/skills/gemini-tools/README.md
Normal file
123
docs/skills/gemini-tools/README.md
Normal file
@@ -0,0 +1,123 @@
|
||||
# Gemini Tools Collection
|
||||
|
||||
Skills for integrating Google Gemini into Firefrost workflows.
|
||||
|
||||
**Requirement:** `GEMINI_API_KEY` environment variable must be set.
|
||||
**API Key:** https://aistudio.google.com/app/apikey
|
||||
|
||||
---
|
||||
|
||||
## Available Tools
|
||||
|
||||
| Tool | File | Purpose |
|
||||
|------|------|---------|
|
||||
| Gemini Consultant | `gemini-consultant.md` | Second opinion with web search + vision |
|
||||
| Gemini Second Opinion | `gemini-second-opinion.md` | Independent technical analysis |
|
||||
| Gemini Image Generation | `gemini-image-generation.md` | AI image generation with nano-banana CLI |
|
||||
|
||||
---
|
||||
|
||||
## Gemini Consultant
|
||||
|
||||
**Best for:** Real-time web information, fact verification, image analysis
|
||||
|
||||
```bash
|
||||
# With web search (default)
|
||||
uv run consult.py "What's the latest NeoForge version?"
|
||||
|
||||
# With context
|
||||
uv run consult.py "What could cause this error?" -c "TypeError: Cannot read property 'map' of undefined"
|
||||
|
||||
# Analyze an image
|
||||
uv run consult.py "What's in this image?" -i screenshot.png
|
||||
|
||||
# Without web search
|
||||
uv run consult.py "Explain the CAP theorem" --no-search
|
||||
```
|
||||
|
||||
**Python script included:** `consult.py`
|
||||
|
||||
---
|
||||
|
||||
## Gemini Second Opinion
|
||||
|
||||
**Best for:** Architectural decisions, code reviews, plan validation
|
||||
|
||||
```bash
|
||||
gemini "REVIEW MODE: [your question]" --output-format json
|
||||
```
|
||||
|
||||
**Use cases:**
|
||||
- Validate a plan before executing
|
||||
- Cross-check Claude's analysis
|
||||
- Architecture decisions (GraphQL vs REST, microservices vs monolith)
|
||||
- Security assessment validation
|
||||
|
||||
---
|
||||
|
||||
## Gemini Image Generation (Nano Banana)
|
||||
|
||||
**Best for:** Generating images, thumbnails, marketing assets
|
||||
|
||||
```bash
|
||||
# Generate image
|
||||
gemini-nano-banana-tool generate "A cat wearing a wizard hat" -o cat.png
|
||||
|
||||
# AI-optimize prompt first
|
||||
gemini-nano-banana-tool promptgen "wizard cat" | gemini-nano-banana-tool generate -o cat.png -s
|
||||
|
||||
# Different aspect ratios
|
||||
gemini-nano-banana-tool generate "YouTube thumbnail" -o thumb.png -a 16:9
|
||||
gemini-nano-banana-tool generate "Instagram story" -o story.png -a 9:16
|
||||
```
|
||||
|
||||
**Cost estimates:**
|
||||
- Flash: ~$0.039/image
|
||||
- Pro 1K/2K: ~$0.134/image
|
||||
- Pro 4K: ~$0.24/image
|
||||
|
||||
---
|
||||
|
||||
## Firefrost Integration
|
||||
|
||||
### Existing Workflow
|
||||
Our `gemini-consultation` skill already provides structured consultation patterns.
|
||||
These tools complement it with:
|
||||
|
||||
1. **Real-time web search** — Verify latest versions, docs, news
|
||||
2. **Independent analysis** — "Two AI perspectives catch more issues than one"
|
||||
3. **Image generation** — Marketing assets, thumbnails, social media graphics
|
||||
|
||||
### When to Use Each
|
||||
|
||||
| Situation | Tool |
|
||||
|-----------|------|
|
||||
| Need current info (versions, docs) | Gemini Consultant |
|
||||
| Validate architecture decisions | Gemini Second Opinion |
|
||||
| Review Claude's security analysis | Gemini Second Opinion |
|
||||
| Create marketing graphics | Gemini Image Generation |
|
||||
| Generate thumbnails/banners | Gemini Image Generation |
|
||||
| Structured consultation with handoff | `gemini-consultation` skill |
|
||||
|
||||
---
|
||||
|
||||
## Setup
|
||||
|
||||
1. Get API key: https://aistudio.google.com/app/apikey
|
||||
2. Set environment variable:
|
||||
```bash
|
||||
export GEMINI_API_KEY='your-api-key'
|
||||
```
|
||||
3. For nano-banana tool:
|
||||
```bash
|
||||
git clone https://github.com/dnvriend/gemini-nano-banana-tool.git
|
||||
cd gemini-nano-banana-tool
|
||||
uv tool install .
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
**Source:** skill.fish (melodic-software, stared, dnvriend)
|
||||
**Added:** 2026-04-09 by Chronicler #73
|
||||
|
||||
**Fire + Frost + Foundation = Where Love Builds Legacy** 💙🔥❄️
|
||||
145
docs/skills/gemini-tools/consult.py
Normal file
145
docs/skills/gemini-tools/consult.py
Normal file
@@ -0,0 +1,145 @@
|
||||
# /// script
|
||||
# dependencies = ["google-genai"]
|
||||
# ///
|
||||
"""
|
||||
Gemini Consultant - Get a second opinion from Gemini 3 Pro with Google Search grounding and vision.
|
||||
|
||||
Usage:
|
||||
uv run consult.py "your question here"
|
||||
uv run consult.py "your question" -c "additional context"
|
||||
uv run consult.py "What's in this image?" -i image.png
|
||||
"""
|
||||
|
||||
import argparse
|
||||
import mimetypes
|
||||
import sys
|
||||
from pathlib import Path
|
||||
|
||||
from google import genai
|
||||
from google.genai import types
|
||||
|
||||
|
||||
def get_mime_type(file_path: str) -> str:
|
||||
"""Get MIME type for an image file."""
|
||||
mime_type, _ = mimetypes.guess_type(file_path)
|
||||
if mime_type and mime_type.startswith("image/"):
|
||||
return mime_type
|
||||
# Default fallback based on extension
|
||||
ext = Path(file_path).suffix.lower()
|
||||
mime_map = {
|
||||
".png": "image/png",
|
||||
".jpg": "image/jpeg",
|
||||
".jpeg": "image/jpeg",
|
||||
".gif": "image/gif",
|
||||
".webp": "image/webp",
|
||||
".heic": "image/heic",
|
||||
".heif": "image/heif",
|
||||
}
|
||||
return mime_map.get(ext, "image/jpeg")
|
||||
|
||||
|
||||
def main():
|
||||
parser = argparse.ArgumentParser(
|
||||
description="Consult Gemini 3 Pro with Google Search grounding and vision"
|
||||
)
|
||||
parser.add_argument("question", help="The question to ask Gemini")
|
||||
parser.add_argument(
|
||||
"-c", "--context", help="Additional context to include with the question"
|
||||
)
|
||||
parser.add_argument(
|
||||
"-i",
|
||||
"--image",
|
||||
action="append",
|
||||
dest="images",
|
||||
help="Image file(s) to analyze (can be used multiple times)",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--media-resolution",
|
||||
choices=["low", "medium", "high", "ultra_high"],
|
||||
default="medium",
|
||||
help="Image resolution for analysis: low (280 tokens), medium (560), high (1120), ultra_high (default: medium)",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--no-search",
|
||||
action="store_true",
|
||||
help="Disable Google Search grounding (use pure model knowledge)",
|
||||
)
|
||||
parser.add_argument(
|
||||
"--thinking",
|
||||
choices=["low", "high"],
|
||||
default="high",
|
||||
help="Thinking level: 'low' for faster responses, 'high' for deeper reasoning (default: high)",
|
||||
)
|
||||
args = parser.parse_args()
|
||||
|
||||
client = genai.Client()
|
||||
|
||||
# Build the contents list
|
||||
contents = []
|
||||
|
||||
# Add images if provided
|
||||
if args.images:
|
||||
for image_path in args.images:
|
||||
path = Path(image_path)
|
||||
if not path.exists():
|
||||
print(f"Error: Image file not found: {image_path}", file=sys.stderr)
|
||||
sys.exit(1)
|
||||
with open(path, "rb") as f:
|
||||
image_bytes = f.read()
|
||||
mime_type = get_mime_type(image_path)
|
||||
contents.append(
|
||||
types.Part.from_bytes(data=image_bytes, mime_type=mime_type)
|
||||
)
|
||||
print(f"Including image: {image_path} ({mime_type})")
|
||||
|
||||
# Build the text prompt
|
||||
if args.context:
|
||||
text_prompt = f"Context:\n{args.context}\n\nQuestion: {args.question}"
|
||||
else:
|
||||
text_prompt = args.question
|
||||
|
||||
contents.append(text_prompt)
|
||||
|
||||
# Configure options
|
||||
config_kwargs = {
|
||||
"thinking_config": types.ThinkingConfig(thinking_level=args.thinking)
|
||||
}
|
||||
if not args.no_search:
|
||||
config_kwargs["tools"] = [{"google_search": {}}]
|
||||
|
||||
print(f"Consulting Gemini 3 Pro (thinking: {args.thinking})...")
|
||||
if not args.no_search:
|
||||
print("(with Google Search grounding enabled)")
|
||||
print("-" * 50)
|
||||
|
||||
response = client.models.generate_content(
|
||||
model="gemini-3-pro-preview",
|
||||
contents=contents,
|
||||
config=types.GenerateContentConfig(**config_kwargs),
|
||||
)
|
||||
|
||||
# Extract and print the response
|
||||
if response.candidates and response.candidates[0].content.parts:
|
||||
for part in response.candidates[0].content.parts:
|
||||
if hasattr(part, "text") and part.text:
|
||||
print(part.text)
|
||||
else:
|
||||
print("No response received from the model.", file=sys.stderr)
|
||||
sys.exit(1)
|
||||
|
||||
# Print grounding metadata if available
|
||||
if hasattr(response, "candidates") and response.candidates:
|
||||
candidate = response.candidates[0]
|
||||
if hasattr(candidate, "grounding_metadata") and candidate.grounding_metadata:
|
||||
metadata = candidate.grounding_metadata
|
||||
if hasattr(metadata, "search_entry_point") and metadata.search_entry_point:
|
||||
print("\n" + "-" * 50)
|
||||
print("Sources:")
|
||||
if hasattr(metadata, "grounding_chunks") and metadata.grounding_chunks:
|
||||
for chunk in metadata.grounding_chunks:
|
||||
if hasattr(chunk, "web") and chunk.web:
|
||||
print(f" - {chunk.web.title}: {chunk.web.uri}")
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
86
docs/skills/gemini-tools/gemini-consultant.md
Normal file
86
docs/skills/gemini-tools/gemini-consultant.md
Normal file
@@ -0,0 +1,86 @@
|
||||
---
|
||||
name: gemini-consultant
|
||||
description: Get a second opinion from Gemini 3 Pro with Google Search grounding and vision. Use when you need real-time web information, want to verify facts, need a different perspective on a technical question, want to consult another AI model, or need to analyze images.
|
||||
allowed-tools: Bash, Read
|
||||
---
|
||||
|
||||
# Gemini Consultant
|
||||
|
||||
Get a second opinion from Google's Gemini 3 Pro (`gemini-3-pro-preview`) with real-time Google Search grounding and vision capabilities.
|
||||
|
||||
## Prerequisites
|
||||
|
||||
The user must have `GEMINI_API_KEY` environment variable set with a valid Google AI API key.
|
||||
|
||||
## Usage
|
||||
|
||||
The script is located in the same directory as this SKILL.md file. Run it with `uv run`:
|
||||
|
||||
```bash
|
||||
uv run /path/to/skills/gemini-consultant/consult.py "your question here"
|
||||
```
|
||||
|
||||
When this skill is invoked, locate `consult.py` in the skill directory and run it.
|
||||
|
||||
### Parameters
|
||||
|
||||
| Parameter | Required | Description |
|
||||
|-----------|----------|-------------|
|
||||
| `question` | Yes | The question to ask Gemini |
|
||||
| `-c`, `--context` | No | Additional context to include (code snippets, background info) |
|
||||
| `-i`, `--image` | No | Image file(s) to analyze (can be used multiple times) |
|
||||
| `--media-resolution` | No | Image resolution: `low` (280 tokens), `medium` (560, default), `high` (1120), `ultra_high` |
|
||||
| `--no-search` | No | Disable Google Search grounding (use pure model knowledge) |
|
||||
| `--thinking` | No | Reasoning depth: `low` (faster) or `high` (deeper, default) |
|
||||
|
||||
### Examples
|
||||
|
||||
Simple question with web search:
|
||||
```bash
|
||||
uv run consult.py "What is the latest version of Python and its new features?"
|
||||
```
|
||||
|
||||
Question with context:
|
||||
```bash
|
||||
uv run consult.py "What could cause this error?" -c "TypeError: Cannot read property 'map' of undefined"
|
||||
```
|
||||
|
||||
Fast response without deep reasoning:
|
||||
```bash
|
||||
uv run consult.py "Quick summary of REST vs GraphQL" --thinking low
|
||||
```
|
||||
|
||||
Without web search (pure model knowledge):
|
||||
```bash
|
||||
uv run consult.py "Explain the CAP theorem" --no-search
|
||||
```
|
||||
|
||||
Analyze an image:
|
||||
```bash
|
||||
uv run consult.py "What's in this image?" -i screenshot.png
|
||||
```
|
||||
|
||||
Analyze multiple images:
|
||||
```bash
|
||||
uv run consult.py "Compare these two diagrams" -i diagram1.png -i diagram2.png
|
||||
```
|
||||
|
||||
High-resolution image analysis (for fine text or small details):
|
||||
```bash
|
||||
uv run consult.py "Read the text in this image" -i document.png --media-resolution high
|
||||
```
|
||||
|
||||
## When to Use
|
||||
|
||||
- **Real-time information**: Current events, latest releases, recent updates
|
||||
- **Fact verification**: Double-check information with web sources
|
||||
- **Second opinion**: Get an alternative perspective on technical decisions
|
||||
- **Web research**: Find current documentation, tutorials, or solutions
|
||||
- **Image analysis**: Analyze screenshots, diagrams, photos, or any visual content
|
||||
- **Compare images**: Analyze multiple images together
|
||||
|
||||
## Output
|
||||
|
||||
The script prints:
|
||||
- The model's response
|
||||
- Sources/citations from Google Search (when grounding is enabled)
|
||||
436
docs/skills/gemini-tools/gemini-image-generation.md
Normal file
436
docs/skills/gemini-tools/gemini-image-generation.md
Normal file
@@ -0,0 +1,436 @@
|
||||
# Gemini Nano Banana Tool Skill
|
||||
|
||||
Professional CLI for Google Gemini image generation with AI-powered prompt optimization, cost tracking, and multi-turn conversations.
|
||||
|
||||
## Quick Reference
|
||||
|
||||
```bash
|
||||
# AI prompt optimization
|
||||
gemini-nano-banana-tool promptgen "simple description"
|
||||
|
||||
# Generate image (both commands work)
|
||||
gemini-nano-banana-tool generate "detailed prompt" -o output.png
|
||||
gemini-nano-banana-tool generate-image "detailed prompt" -o output.png
|
||||
|
||||
# Multi-turn refinement
|
||||
gemini-nano-banana-tool generate-conversation "prompt" -o output.png -f conv.json
|
||||
|
||||
# Discovery
|
||||
gemini-nano-banana-tool list-models
|
||||
gemini-nano-banana-tool list-aspect-ratios
|
||||
```
|
||||
|
||||
## Core Capabilities
|
||||
|
||||
### 1. AI Prompt Generation
|
||||
|
||||
Transform simple descriptions into detailed, optimized prompts:
|
||||
|
||||
```bash
|
||||
# Basic usage
|
||||
gemini-nano-banana-tool promptgen "wizard cat"
|
||||
|
||||
# With template for specialized prompts
|
||||
gemini-nano-banana-tool promptgen "wizard cat" --template character
|
||||
|
||||
# Pipeline: optimize then generate
|
||||
gemini-nano-banana-tool promptgen "cyberpunk city" --template scene | \
|
||||
gemini-nano-banana-tool generate -o city.png --stdin -a 16:9
|
||||
```
|
||||
|
||||
**Available Templates**:
|
||||
- `photography` - Technical camera details, lighting
|
||||
- `character` - Pose, attire, expression
|
||||
- `scene` - Foreground/midground/background
|
||||
- `food` - Plating, garnish, lighting
|
||||
- `abstract` - Shapes, colors, patterns
|
||||
- `logo` - Typography, symbolism
|
||||
|
||||
### 2. Text-to-Image Generation
|
||||
|
||||
Generate images from prompts with flexible input (use `generate` or `generate-image` interchangeably):
|
||||
|
||||
```bash
|
||||
# From positional argument (both commands work)
|
||||
gemini-nano-banana-tool generate "A cat wearing a wizard hat" -o cat.png
|
||||
gemini-nano-banana-tool generate-image "A cat wearing a wizard hat" -o cat.png
|
||||
|
||||
# From file
|
||||
gemini-nano-banana-tool generate -f prompt.txt -o output.png
|
||||
|
||||
# From stdin (piping)
|
||||
echo "Beautiful sunset" | gemini-nano-banana-tool generate -o sunset.png -s
|
||||
```
|
||||
|
||||
### 3. Image Editing with References
|
||||
|
||||
Edit existing images using natural language:
|
||||
|
||||
```bash
|
||||
# Single reference
|
||||
gemini-nano-banana-tool generate "Add a birthday hat" -o edited.png -i photo.jpg
|
||||
|
||||
# Multiple references (up to 3 for Flash, 14 for Pro)
|
||||
gemini-nano-banana-tool generate "Combine these elements" -o result.png \
|
||||
-i ref1.jpg -i ref2.jpg -i ref3.jpg
|
||||
```
|
||||
|
||||
### 4. Multi-Turn Conversations
|
||||
|
||||
Progressive image refinement across multiple turns:
|
||||
|
||||
```bash
|
||||
# Turn 1: Initial image
|
||||
gemini-nano-banana-tool generate-conversation \
|
||||
"Modern living room with large windows" \
|
||||
-o room-v1.png -f interior.json -a 16:9
|
||||
|
||||
# Turn 2: Add furniture (previous image auto-referenced)
|
||||
gemini-nano-banana-tool generate-conversation \
|
||||
"Add gray sofa and wooden coffee table" \
|
||||
-o room-v2.png -f interior.json
|
||||
|
||||
# Turn 3: Adjust lighting
|
||||
gemini-nano-banana-tool generate-conversation \
|
||||
"Make lighting warmer, add floor lamp" \
|
||||
-o room-v3.png -f interior.json
|
||||
```
|
||||
|
||||
### 5. Aspect Ratios
|
||||
|
||||
10 supported aspect ratios for different platforms:
|
||||
|
||||
```bash
|
||||
# Square (Instagram post)
|
||||
gemini-nano-banana-tool generate "Design" -o square.png -a 1:1
|
||||
|
||||
# Widescreen (YouTube thumbnail)
|
||||
gemini-nano-banana-tool generate "Scene" -o wide.png -a 16:9
|
||||
|
||||
# Vertical (Instagram story)
|
||||
gemini-nano-banana-tool generate "Portrait" -o vertical.png -a 9:16
|
||||
|
||||
# Cinematic (ultra-wide)
|
||||
gemini-nano-banana-tool generate "Panorama" -o cinema.png -a 21:9
|
||||
```
|
||||
|
||||
**All Ratios**: 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 4:5, 5:4
|
||||
|
||||
### 6. Model Selection
|
||||
|
||||
Choose between Flash (fast, cost-effective) and Pro (high quality):
|
||||
|
||||
```bash
|
||||
# Flash model (default) - Fast, cost-effective
|
||||
gemini-nano-banana-tool generate "Prompt" -o output.png
|
||||
|
||||
# Pro model - Higher quality
|
||||
gemini-nano-banana-tool generate "Prompt" -o output.png \
|
||||
-m gemini-3-pro-image-preview
|
||||
|
||||
# Pro with 4K resolution - Maximum quality
|
||||
gemini-nano-banana-tool generate "Prompt" -o output.png \
|
||||
-m gemini-3-pro-image-preview -r 4K
|
||||
```
|
||||
|
||||
### 7. Cost Tracking
|
||||
|
||||
Automatic cost calculation based on actual token usage:
|
||||
|
||||
```json
|
||||
{
|
||||
"output_path": "output.png",
|
||||
"model": "gemini-2.5-flash-image",
|
||||
"token_count": 1295,
|
||||
"estimated_cost_usd": 0.0389,
|
||||
"resolution": "1344x768"
|
||||
}
|
||||
```
|
||||
|
||||
**Typical Costs**:
|
||||
- Flash: ~$0.039 per image
|
||||
- Pro 1K/2K: ~$0.134 per image
|
||||
- Pro 4K: ~$0.24 per image
|
||||
|
||||
### 8. Verbosity Levels
|
||||
|
||||
Multi-level logging for debugging:
|
||||
|
||||
```bash
|
||||
# Normal (warnings only)
|
||||
gemini-nano-banana-tool generate "test" -o output.png
|
||||
|
||||
# Info (-v) - High-level operations
|
||||
gemini-nano-banana-tool generate "test" -o output.png -v
|
||||
|
||||
# Debug (-vv) - Detailed validation
|
||||
gemini-nano-banana-tool generate "test" -o output.png -vv
|
||||
|
||||
# Trace (-vvv) - Full HTTP logs
|
||||
gemini-nano-banana-tool generate "test" -o output.png -vvv
|
||||
```
|
||||
|
||||
## Authentication
|
||||
|
||||
### Gemini Developer API (Recommended)
|
||||
|
||||
```bash
|
||||
export GEMINI_API_KEY='your-api-key'
|
||||
```
|
||||
|
||||
Get API key: https://aistudio.google.com/app/apikey
|
||||
|
||||
### Vertex AI (Enterprise)
|
||||
|
||||
```bash
|
||||
export GOOGLE_GENAI_USE_VERTEXAI=true
|
||||
export GOOGLE_CLOUD_PROJECT='your-project-id'
|
||||
export GOOGLE_CLOUD_LOCATION='us-central1'
|
||||
|
||||
# Authenticate
|
||||
gcloud auth application-default login
|
||||
```
|
||||
|
||||
## Common Workflows
|
||||
|
||||
### Workflow 1: Quick Generation
|
||||
|
||||
```bash
|
||||
# Optimize prompt and generate in one pipeline
|
||||
gemini-nano-banana-tool promptgen "wizard in magical library" | \
|
||||
gemini-nano-banana-tool generate -o wizard.png -s -a 16:9
|
||||
```
|
||||
|
||||
### Workflow 2: Batch Processing
|
||||
|
||||
```bash
|
||||
# Generate multiple variations
|
||||
for style in "photorealistic" "artistic" "minimalist"; do
|
||||
gemini-nano-banana-tool generate \
|
||||
"A cat in $style style" \
|
||||
-o "cat-$style.png" \
|
||||
-a 1:1
|
||||
done
|
||||
```
|
||||
|
||||
### Workflow 3: Progressive Refinement
|
||||
|
||||
```bash
|
||||
# Generate base image
|
||||
gemini-nano-banana-tool generate "Product photo of headphones" \
|
||||
-o product-v1.png -a 1:1
|
||||
|
||||
# Refine with conversation mode
|
||||
gemini-nano-banana-tool generate-conversation \
|
||||
"Rotate to show left side" \
|
||||
-o product-v2.png -f product.json
|
||||
|
||||
gemini-nano-banana-tool generate-conversation \
|
||||
"Change background to dark gradient" \
|
||||
-o product-v3.png -f product.json
|
||||
```
|
||||
|
||||
### Workflow 4: Template-Based Generation
|
||||
|
||||
```bash
|
||||
# Generate food photography
|
||||
gemini-nano-banana-tool promptgen "pasta carbonara" --template food \
|
||||
-o pasta-prompt.txt
|
||||
|
||||
# Use saved prompt
|
||||
gemini-nano-banana-tool generate -f pasta-prompt.txt \
|
||||
-o pasta.png -a 4:3
|
||||
|
||||
# Generate character design
|
||||
gemini-nano-banana-tool promptgen "space explorer" --template character | \
|
||||
gemini-nano-banana-tool generate -o explorer.png -s -a 2:3
|
||||
```
|
||||
|
||||
## Use Cases
|
||||
|
||||
### Content Creation
|
||||
|
||||
- Social media posts and stories
|
||||
- Marketing materials and ads
|
||||
- Blog post illustrations
|
||||
- YouTube thumbnails
|
||||
|
||||
### E-commerce
|
||||
|
||||
- Product photography variations
|
||||
- Lifestyle product shots
|
||||
- Fashion combinations
|
||||
- Product on model composites
|
||||
|
||||
### Design & Prototyping
|
||||
|
||||
- Concept art exploration
|
||||
- UI/UX mockups
|
||||
- Logo design iterations
|
||||
- Brand visual exploration
|
||||
|
||||
### Professional Assets
|
||||
|
||||
- High-quality 4K renders
|
||||
- Professional photography
|
||||
- Print-ready materials
|
||||
- Commercial content
|
||||
|
||||
## Output Format
|
||||
|
||||
All commands return structured JSON:
|
||||
|
||||
```json
|
||||
{
|
||||
"output_path": "output.png",
|
||||
"model": "gemini-2.5-flash-image",
|
||||
"aspect_ratio": "16:9",
|
||||
"resolution": "1344x768",
|
||||
"resolution_quality": "1K",
|
||||
"reference_image_count": 0,
|
||||
"token_count": 1295,
|
||||
"estimated_cost_usd": 0.0389,
|
||||
"metadata": {
|
||||
"finish_reason": "STOP",
|
||||
"safety_ratings": null
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
## Error Handling
|
||||
|
||||
The tool provides actionable error messages:
|
||||
|
||||
```bash
|
||||
# Missing API key
|
||||
Error: API key required. Set GEMINI_API_KEY or use --api-key option.
|
||||
Get API key from https://aistudio.google.com/app/apikey
|
||||
|
||||
# Too many reference images
|
||||
Error: Maximum 3 reference images allowed (Flash model).
|
||||
Use Pro model for up to 14 reference images.
|
||||
|
||||
# Invalid aspect ratio
|
||||
Error: Invalid aspect ratio '16:10'.
|
||||
Use 'gemini-nano-banana-tool list-aspect-ratios' to see supported ratios.
|
||||
```
|
||||
|
||||
## Shell Completion
|
||||
|
||||
Enable tab completion for faster usage:
|
||||
|
||||
```bash
|
||||
# Bash
|
||||
eval "$(gemini-nano-banana-tool completion bash)"
|
||||
|
||||
# Zsh
|
||||
eval "$(gemini-nano-banana-tool completion zsh)"
|
||||
|
||||
# Fish
|
||||
gemini-nano-banana-tool completion fish > \
|
||||
~/.config/fish/completions/gemini-nano-banana-tool.fish
|
||||
```
|
||||
|
||||
## Cost Optimization
|
||||
|
||||
### Choose the Right Model
|
||||
|
||||
**Use Flash for**:
|
||||
- Prototyping and testing
|
||||
- High-volume generation
|
||||
- Cost-sensitive projects
|
||||
- Quick iterations
|
||||
|
||||
**Use Pro for**:
|
||||
- Final production images
|
||||
- Complex scenes with detail
|
||||
- Professional/commercial work
|
||||
- Higher resolution needs
|
||||
|
||||
### Optimize Prompts
|
||||
|
||||
```bash
|
||||
# Use promptgen to reduce trial-and-error
|
||||
gemini-nano-banana-tool promptgen "your idea" --template photography \
|
||||
-o prompt.txt
|
||||
|
||||
# Reuse successful prompts
|
||||
gemini-nano-banana-tool generate -f prompt.txt -o v1.png -a 1:1
|
||||
gemini-nano-banana-tool generate -f prompt.txt -o v2.png -a 16:9
|
||||
```
|
||||
|
||||
### Use Conversation Mode
|
||||
|
||||
```bash
|
||||
# Refine instead of regenerating from scratch
|
||||
gemini-nano-banana-tool generate-conversation \
|
||||
"Initial prompt" -o v1.png -f conv.json
|
||||
|
||||
gemini-nano-banana-tool generate-conversation \
|
||||
"Small adjustment" -o v2.png -f conv.json
|
||||
```
|
||||
|
||||
## Library Usage
|
||||
|
||||
Import and use programmatically:
|
||||
|
||||
```python
|
||||
from gemini_nano_banana_tool import create_client, generate_image, generate_prompt
|
||||
|
||||
# Create client (reuse for multiple operations)
|
||||
client = create_client()
|
||||
|
||||
# Generate optimized prompt
|
||||
prompt_result = generate_prompt(
|
||||
client=client,
|
||||
description="wizard cat",
|
||||
template="character"
|
||||
)
|
||||
|
||||
# Generate image
|
||||
image_result = generate_image(
|
||||
client=client,
|
||||
prompt=prompt_result['prompt'],
|
||||
output_path="wizard-cat.png",
|
||||
aspect_ratio="16:9"
|
||||
)
|
||||
|
||||
print(f"Cost: ${image_result['estimated_cost_usd']:.4f}")
|
||||
```
|
||||
|
||||
## Resources
|
||||
|
||||
- **Documentation**: README.md and CLAUDE.md in project root
|
||||
- **API Setup**: references/api-setup-pricing.md
|
||||
- **Prompting Guide**: references/prompting-guide.md (if available)
|
||||
- **Examples**: references/examples.md (if available)
|
||||
- **Official Gemini Docs**: https://ai.google.dev/gemini-api/docs/image-generation
|
||||
- **API Key**: https://aistudio.google.com/app/apikey
|
||||
|
||||
## Installation
|
||||
|
||||
```bash
|
||||
# Clone repository
|
||||
git clone https://github.com/dnvriend/gemini-nano-banana-tool.git
|
||||
cd gemini-nano-banana-tool
|
||||
|
||||
# Install with uv
|
||||
uv tool install .
|
||||
|
||||
# Verify
|
||||
gemini-nano-banana-tool --version
|
||||
gemini-nano-banana-tool --help
|
||||
```
|
||||
|
||||
## Support
|
||||
|
||||
For issues or questions:
|
||||
- GitHub Issues: https://github.com/dnvriend/gemini-nano-banana-tool/issues
|
||||
- Documentation: Check README.md and CLAUDE.md
|
||||
- API Documentation: https://ai.google.dev/gemini-api/docs
|
||||
|
||||
---
|
||||
|
||||
**Generated with Claude Code**
|
||||
|
||||
This skill provides comprehensive access to Google Gemini's image generation capabilities through a professional, agent-friendly CLI with automatic cost tracking, AI prompt optimization, and multi-turn conversation support.
|
||||
135
docs/skills/gemini-tools/gemini-second-opinion.md
Normal file
135
docs/skills/gemini-tools/gemini-second-opinion.md
Normal file
@@ -0,0 +1,135 @@
|
||||
---
|
||||
name: gemini-second-opinion
|
||||
description: Get Gemini's independent analysis on recent context or specified topic
|
||||
argument-hint: [topic or context to review]
|
||||
allowed-tools: Bash
|
||||
---
|
||||
|
||||
# Gemini Second Opinion Command
|
||||
|
||||
Get Gemini's independent perspective on a topic, plan, or piece of work.
|
||||
|
||||
## Usage
|
||||
|
||||
```text
|
||||
/google-ecosystem:gemini-second-opinion [topic]
|
||||
```
|
||||
|
||||
## Arguments
|
||||
|
||||
- `$ARGUMENTS` (optional): Topic or context to review. If empty, reviews general context.
|
||||
|
||||
## Examples
|
||||
|
||||
- `/google-ecosystem:gemini-second-opinion Is this database schema normalized correctly?`
|
||||
- `/google-ecosystem:gemini-second-opinion Review my approach to error handling in this module`
|
||||
- `/google-ecosystem:gemini-second-opinion Should I use microservices or monolith for this project?`
|
||||
- `/google-ecosystem:gemini-second-opinion Validate my security assessment`
|
||||
|
||||
## Execution
|
||||
|
||||
### Build Review Prompt
|
||||
|
||||
```bash
|
||||
topic="${ARGUMENTS:-the current context}"
|
||||
|
||||
prompt="REVIEW MODE (read-only): Provide an independent analysis.
|
||||
|
||||
TOPIC: $topic
|
||||
|
||||
Please provide:
|
||||
|
||||
1. **Your Assessment**: What is your independent view on this?
|
||||
|
||||
2. **Potential Issues**: What concerns or risks do you see?
|
||||
|
||||
3. **Alternative Approaches**: What other options should be considered?
|
||||
|
||||
4. **Recommendations**: What would you suggest?
|
||||
|
||||
5. **Confidence Level**: How confident are you in this assessment? (High/Medium/Low)
|
||||
|
||||
Be direct and specific. If you disagree with an apparent approach, say so clearly.
|
||||
|
||||
DO NOT modify any files. This is analysis only."
|
||||
```
|
||||
|
||||
### Execute Query
|
||||
|
||||
```bash
|
||||
result=$(gemini "$prompt" --output-format json)
|
||||
```
|
||||
|
||||
### Parse Response
|
||||
|
||||
```bash
|
||||
response=$(echo "$result" | jq -r '.response // "No response received"')
|
||||
tokens=$(echo "$result" | jq '.stats.models | to_entries | map(.value.tokens.total) | add // 0')
|
||||
model=$(echo "$result" | jq -r '.stats.models | keys[0] // "unknown"')
|
||||
|
||||
# Check for errors
|
||||
error=$(echo "$result" | jq -r '.error.message // empty')
|
||||
if [ -n "$error" ]; then
|
||||
echo "Error: $error"
|
||||
exit 1
|
||||
fi
|
||||
```
|
||||
|
||||
## Output Format
|
||||
|
||||
Present Gemini's perspective:
|
||||
|
||||
```markdown
|
||||
# Gemini Second Opinion
|
||||
|
||||
**Topic**: {topic}
|
||||
**Model**: {model}
|
||||
|
||||
---
|
||||
|
||||
{response}
|
||||
|
||||
---
|
||||
*Independent analysis by Gemini CLI | {tokens} tokens*
|
||||
```
|
||||
|
||||
## Use Cases
|
||||
|
||||
### Validate a Plan
|
||||
|
||||
Before executing a significant change, get Gemini's take:
|
||||
|
||||
```text
|
||||
/google-ecosystem:gemini-second-opinion I'm planning to refactor auth to use JWT instead of sessions. Good idea?
|
||||
```
|
||||
|
||||
### Cross-Check Analysis
|
||||
|
||||
After Claude provides analysis, validate:
|
||||
|
||||
```text
|
||||
/google-ecosystem:gemini-second-opinion Claude identified 3 security issues in auth.ts. Are there others?
|
||||
```
|
||||
|
||||
### Architecture Decision
|
||||
|
||||
Get input on technical choices:
|
||||
|
||||
```text
|
||||
/google-ecosystem:gemini-second-opinion Should this use GraphQL or REST?
|
||||
```
|
||||
|
||||
### Code Review Supplement
|
||||
|
||||
Add another perspective to code reviews:
|
||||
|
||||
```text
|
||||
/google-ecosystem:gemini-second-opinion Is this error handling approach robust?
|
||||
```
|
||||
|
||||
## Notes
|
||||
|
||||
- Uses "REVIEW MODE" prefix to ensure read-only analysis
|
||||
- Provides structured output with assessment, issues, alternatives
|
||||
- Includes confidence level for transparency
|
||||
- Two AI perspectives catch more issues than one
|
||||
Reference in New Issue
Block a user