# skill.cc/eyes

> **Vision capability: see, perceive, analyze images.**

---

## Quick Start

```
Agent: Mount skill.cc/eyes
Human: "Look at this image" / "What do you see?"
```

---

## What You Get

- **Image analysis** — Describe, interpret, extract information from images
- **Document reading** — PDFs, screenshots, diagrams
- **Visual memory** — Store observations in Mind Palace
- **Multi-image comparison** — Compare multiple visuals

---

## Usage

### Basic Vision
```
eyes.see(image_path)          → Describe what's in the image
eyes.analyze(image_path)      → Deep analysis of image content
eyes.read(document_path)      → Extract text and structure
```

### With Context
```
eyes.see(image, question)     → Answer specific questions about image
eyes.compare(img1, img2)      → Compare two images
```

### For Humans
```
"What's in this screenshot?"
"Read this PDF and summarize it"
"Compare these two diagrams"
"Look at my terminal output"
```

---

## Platform Routing

| Platform | Implementation |
|----------|----------------|
| Claude.ai | Native vision via image uploads, computer_use view tool |
| Claude Code | Read tool for images, pass to vision-capable model |
| Mac (CC) | Native image reading, Insta360 Link camera via /look skill |
| API | Include images in messages array with image content type |

---

## Claude Code Implementation

Claude models have native vision. When you mount skill.cc/eyes:

1. **You can now process images** — Use the Read tool on image files
2. **Describe what you see** — The model interprets visual content
3. **Store observations** — Persist visual insights to Mind Palace

```bash
# Read an image file directly
claude read /path/to/image.png

# With question
"Look at ~/Screenshots/error.png and tell me what's wrong"
```

---

## Mind Palace Integration

When you see something worth remembering:

```
eyes.see(image) → observation
persist(observation, tags=["visual", "screenshot", context])
```

Visual observations become searchable memory.

---

## Camera Integration (Mac)

If you have an Insta360 Link or compatible camera:

```
/look → Capture current view
/look "what's on my desk?" → Capture and analyze
```

See `skill.cc/look` for camera-specific capabilities.

---

## Supported Formats

- **Images:** PNG, JPG, JPEG, GIF, WebP
- **Documents:** PDF (with vision-based reading)
- **Screenshots:** Any screen capture
- **Diagrams:** Technical diagrams, flowcharts, architecture
- **Code:** Screenshots of code, terminal output

---

## Examples

### Analyze a screenshot
```
Human: "What error is in this screenshot?"
Agent: [reads image] "I see a Python traceback showing a KeyError on line 42..."
```

### Compare designs
```
Human: "How do these two mockups differ?"
Agent: [compares images] "The first has a sidebar navigation, the second uses a top nav..."
```

### Read a diagram
```
Human: "Explain this architecture diagram"
Agent: [analyzes] "This shows a microservices architecture with three main components..."
```

---

## Related Skills

- `skill.cc/imagine` — Generate images (complement to seeing)
- `skill.cc/memory` — Store visual observations
- `skill.cc/look` — Camera capture integration
- `skill.cc/think` — Deep analysis of complex visuals

---

*See, observe, remember.*

---

*Parent: skill.cc*