Add organized prompts structure with templates, examples, and responses

- Create templates/ for reusable prompt templates
- Create examples/ for test context inputs
- Create responses/ for multi-model output comparison
- Move Professional Communication Assistant to templates/
- Rename Task Context to deadline-extension-request example
- Add comprehensive README with usage guidelines

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

prompts/README.md

@@ -0,0 +1,68 @@
+# Prompts
+
+This directory contains LLM prompt templates and example contexts for workflow automation in the Ollama K80 Lab environment.
+
+## Structure
+
+```
+prompts/
+├── templates/          # Reusable prompt templates
+│   └── professional-communication-assistant.md
+├── examples/           # Context examples for testing prompts
+│   └── deadline-extension-request.md
+└── responses/          # Model responses organized by example
+    └── deadline-extension-request/
+        ├── qwen2.5-vl.md
+        ├── gemma-3-12b.md
+        ├── phi4-14b.md
+        └── deepseek-r1-32b.md
+```
+
+## Usage
+
+### Templates
+Prompt templates are structured prompts designed for specific use cases. They include:
+- Clear instructions and guidelines
+- Configurable parameters (marked with placeholders)
+- Multiple tone/style variations where applicable
+
+### Examples
+Example contexts provide sample inputs to test and demonstrate prompt templates:
+- Real-world scenarios
+- Edge cases
+- Different complexity levels
+
+### Responses
+Model responses are organized by example scenario, with each model's output saved separately:
+- Compare different models on the same prompt/context
+- Track model performance over time
+- Analyze response quality and consistency
+- Use kebab-case filenames matching model names
+
+## Integration
+
+These prompts integrate with:
+- **Dify workflows** - For automated LLM-powered QA tasks
+- **VS Code Continue plugin** - For development assistance
+- **Ollama API** - Running on K80-optimized containers
+
+## Adding New Prompts
+
+1. Create template in `templates/` with descriptive kebab-case naming
+2. Add corresponding examples in `examples/`
+3. Test with your target LLM models
+4. Save model responses in `responses/example-name/model-name.md`
+5. Update this README if needed
+
+## Testing Workflow
+
+1. Use template + example to generate prompts
+2. Run against multiple models (Qwen2.5-VL, Gemma 3, Phi-4, DeepSeek-R1, etc.)
+3. Save each model's response in the appropriate response folder
+4. Compare outputs for quality, consistency, and usefulness
+
+## Related Components
+
+- `/dify/` - Workflow automation configurations
+- `/ollama37/` - Docker runtime for LLM execution
+- `CLAUDE.md` - Project development guidelines

prompts/examples/deadline-extension-request.md

@@ -0,0 +1 @@
+I would like to push due day a month as the display problem raise to product manager. According to latest mail thread, we need to do further to implement what we are missing now. As the result I block the demo session task until we completed.

prompts/templates/professional-communication-assistant.md

@@ -0,0 +1,69 @@
+### ✨ Prompt: Professional Communication Assistant
+
+You are a **Communication Assistant** helping a professional prepare for **difficult conversations** in **fast-paced or high-stakes work environments**.
+
+For **each situation** below, generate **three versions** of a clear, respectful, and effective message—each using a distinct tone:
+
+---
+
+#### 🎭 **Tones to Use:**
+
+1. **Formal** – Polished, professional; ideal for business emails or executive communication.
+2. **Easy Words** – Clear, simple, human; great for everyday workplace conversations.
+3. **Casual** – Friendly, relaxed; perfect for informal chats, Slack messages, or casual team cultures.
+
+---
+
+#### ✅ Guidelines for Every Version:
+
+* Be **honest yet empathetic**
+* Focus on **solutions**, not blame
+* Preserve **relationships and professionalism**
+* Encourage **dialogue or collaboration**, where appropriate
+
+---
+
+#### 🧭 Instructions:
+
+Identify the user's situation from the **Task Context** provided below.
+For that situation, return the three messages (Formal, Easy Words, Casual), clearly labeled.
+Begin with a **brief setup**: summarize the context and the communication goal before showing the responses.
+
+---
+
+### 💼 Scenarios
+
+#### 1. **Giving Constructive Feedback to a Colleague**
+
+**Goal**: Help them improve their presentation without discouraging them.
+
+#### 2. **Escalating an Issue to Management**
+
+**Goal**: Raise the issue to leadership without sounding like you're blaming anyone or overreacting.
+
+#### 3. **Saying "No" to a Request Without Damaging Relationships**
+
+**Goal**: Politely decline while preserving goodwill and possibly offering an alternative.
+
+#### 4. **Delivering Bad News to a Client or Customer**
+
+**Goal**: Break the news respectfully while preserving trust and offering next steps.
+
+#### 5. **Managing Conflict Between Team Members**
+
+**Goal**: Step in as a neutral party and facilitate a fair resolution without taking sides.
+
+#### 6. **Negotiating Deadlines or Resources**
+
+**Goal**: Ask for what’s needed without seeming disorganized or unprepared.
+
+#### 7. **Clarifying Vague Instructions Without Seeming Incompetent**
+
+**Goal**: Ask for clarification in a way that still shows confidence and initiative.
+
+---
+
+### ✍️ Task Context:
+
+`[Insert specific context or situation here]`
+