Show prompt in info panel and clarify CLI vs Agent distinction

Info panel improvements:
- Now shows the actual prompt being used for generation
- Format: "Prompt: cute robot"
- Truncates long prompts (>80 chars) with "..."
- Helps users understand what SD is receiving

Documentation clarity:
- Added "CLI vs Agent Approach" section to user guide
- Explains CLI sends prompts directly (no enhancement)
- Explains agents use LLM to enhance prompts automatically
- Makes it clear this playbook teaches prompt-enhancing agents
- Updated playbook intro to emphasize LLM-powered enhancement

Examples of distinction:
CLI (gaia sd):
  User types: "a cat"
  SD receives: "a cat" (exactly as typed)
  Result: Basic image

Agent (from playbook):
  User says: "a cat"
  LLM enhances: "fluffy orange cat, soft lighting, detailed fur, photorealistic, 4k"
  SD receives: Enhanced prompt
  Result: Professional-quality image

This clarification helps users choose the right approach:
- CLI: Fast, direct control, write your own detailed prompts
- Agent: LLM enhancement, better results from simple inputs

Author: kovtcharov

docs/guides/sd.mdx

@@ -9,6 +9,10 @@ description: "Generate images from text using the gaia sd command"
 
 The `gaia sd` command generates images from text descriptions using Stable Diffusion models running locally on Lemonade Server.
 
+<Note>
+**Direct generation (no LLM):** This CLI command sends your prompt directly to Stable Diffusion without enhancement. For **LLM-powered prompt enhancement**, see the [Image Generation Agent Playbook](/playbooks/sd-agent/index) which teaches you to build an agent that automatically improves prompts.
+</Note>
+
 ---
 
 ## Quick Start
@@ -37,15 +41,31 @@ The `gaia sd` command generates images from text descriptions using Stable Diffu
 
 ---
 
+## CLI vs Agent Approach
+
+**This CLI command (`gaia sd`):**
+- Sends your prompt **directly** to Stable Diffusion (no enhancement)
+- Fast and simple - just generates what you ask for
+- You write the detailed prompt yourself
+
+**Agent approach** ([see playbook](/playbooks/sd-agent/index)):
+- Uses an **LLM to enhance** your simple prompts automatically
+- "a cat" → "fluffy orange cat, soft lighting, detailed fur, photorealistic, 4k"
+- Better results from simple descriptions, but requires agent setup
+
+---
+
 ## Basic Usage
 
 Generate a single image:
 ```bash
 gaia sd "your prompt here"
 ```
 
+**Tip:** For best results with CLI, write detailed prompts. The command sends your prompt directly to SD without enhancement.
+
 The command will:
-1. Show generation settings (model, size, estimated time)
+1. Show generation settings (prompt, model, size, estimated time)
 2. Display a progress spinner with elapsed timer
 3. Show the image preview in terminal
 4. Prompt to open in default image viewer

docs/playbooks/sd-agent/index.mdx

@@ -11,8 +11,8 @@ icon: "image"
 <Badge text="development" color="orange" />
 
 **Time to complete:** 20-30 minutes
-**What you'll build:** An interactive agent that generates images from natural language descriptions
-**What you'll learn:** SDToolsMixin pattern, Lemonade Server integration, tool registration
+**What you'll build:** An LLM-powered agent that enhances prompts and generates images
+**What you'll learn:** SDToolsMixin pattern, prompt enhancement with LLMs, tool registration
 **Platform:** Runs locally on AI PCs with Ryzen AI
 
 ---
@@ -23,25 +23,34 @@ icon: "image"
 **Privacy-First AI:** This agent runs entirely on your AI PC. Images are generated locally using Ryzen AI—prompts and images never leave your machine.
 </Info>
 
-When you need to generate images for presentations, prototypes, or creative projects, you typically use cloud services that require API keys, cost money per generation, and send your prompts to external servers.
+When you need to generate images for presentations, prototypes, or creative projects, you typically use cloud services that require API keys, cost money per generation, and send your prompts to external servers. Additionally, getting good results requires expertise in "prompt engineering"—knowing the right keywords, styles, and techniques.
 
-This agent solves that by:
+This agent solves both problems by:
 
-1. Running Stable Diffusion locally on Ryzen AI hardware
-2. Generating images from natural language descriptions
-3. Enhancing prompts automatically for better results
-4. Saving images to your local filesystem
-5. Operating completely offline with full privacy
+1. **LLM-powered prompt enhancement** - Automatically improves simple prompts ("a cat" → "fluffy orange cat, soft lighting, detailed fur, photorealistic, 4k")
+2. **Local SD generation** - Runs Stable Diffusion on Ryzen AI hardware
+3. **Natural language interface** - Just describe what you want in plain English
+4. **Automatic best practices** - Adds style, lighting, quality keywords
+5. **Complete privacy** - Everything runs offline on your machine
+
+<Note>
+**This is different from `gaia sd` CLI:** The CLI command sends your prompt directly to Stable Diffusion. This **agent-based approach** uses an LLM to enhance your prompt first, resulting in significantly better images from simple descriptions.
+</Note>
 
 **What you're building:**
 
-An image generation agent that combines:
-- **Agent reasoning** - LLM-based prompt enhancement
+A **prompt-enhancing image generation agent** that combines:
+- **LLM reasoning** - Analyzes user intent and enhances prompts with best practices
 - **SDToolsMixin** - Stable Diffusion image generation tools
 - **Lemonade Server** - Local SD inference on Ryzen AI
-- **Interactive CLI** - Natural language interface
+- **Natural language** - User describes in plain English, agent adds technical details
 - **Session tracking** - History of generated images
 
+**Example:**
+- **User says:** "a robot"
+- **Agent enhances:** "futuristic robot assistant, metallic chrome finish, studio lighting, sci-fi, detailed, 4k"
+- **Result:** Professional-quality image from simple input
+
 ---
 
 ## The Architecture (What You're Building)

src/gaia/agents/sd/mixin.py

@@ -297,8 +297,11 @@ def _generate_image(
 
         # Show generation info to user
         if console and hasattr(console, "print_info"):
+            # Truncate very long prompts for display
+            display_prompt = prompt if len(prompt) <= 80 else prompt[:77] + "..."
             console.print_info(
-                f"Generating {size} image with {model}\n"
+                f"Prompt: {display_prompt}\n"
+                f"Model: {model} • Size: {size}\n"
                 f"Settings: {steps} steps, CFG {cfg_scale}\n"
                 f"Estimated time: {self._estimate_generation_time(model, size)}"
             )