AgentDoG Security Agent (GGUF)

A quantized security-focused AI agent model based on Qwen3-4B, optimized for safety and security tasks. This skill provides guidance on deploying and using the AgentDoG-FG-Qwen3-4B model in GGUF format for local, privacy-preserving AI operations.

What This Skill Does

This skill helps you select, download, and deploy the appropriate quantization of the AgentDoG-FG-Qwen3-4B model for security and safety-related AI agent tasks. The model is available in multiple GGUF quantization levels, allowing you to balance quality, speed, and resource requirements based on your specific needs.

Instructions

When a user requests assistance with the AgentDoG security model or asks about local security-focused AI deployment:

1. **Assess Requirements**: Determine the user's hardware constraints, quality needs, and use case specificity.

2. **Recommend Quantization Level**:

- For most users: Recommend **Q4_K_M** (2.8GB) as the optimal balance of quality, speed, and size

- For limited resources: Recommend **IQ3_S** (2.2GB) which beats Q3_K variants

- For maximum quality: Recommend **Q6_K** (3.7GB) for near-original quality

- For extreme resource constraints: Recommend **IQ2_S** or **IQ2_M** (1.7-1.8GB)

- For minimal viable deployment: **IQ1_M** (1.4GB) but warn about quality degradation

3. **Provide Download Instructions**:

```bash

# Example for Q4_K_M (recommended)

wget https://huggingface.co/mradermacher/AgentDoG-FG-Qwen3-4B-i1-GGUF/resolve/main/AgentDoG-FG-Qwen3-4B.i1-Q4_K_M.gguf

```

4. **Explain Usage**:

- Guide users to use llama.cpp, ollama, or other GGUF-compatible inference engines

- Explain the model is specifically trained for security and safety tasks

- Note the model is based on Qwen3-4B with Apache 2.0 license

- Emphasize privacy benefits of local deployment

5. **Quality Guidance**:

- IQ (Importance Matrix) quants generally provide better quality than standard quants at similar sizes

- Higher numbers (Q6 > Q5 > Q4) indicate higher quality but larger file size

- Reference the quantization comparison graph for quality/size tradeoffs

6. **Security Use Cases**:

- Vulnerability assessment assistance

- Security policy analysis

- Threat modeling support

- Safety-critical system evaluation

- Offline security research (no data leaves local machine)

Example Usage

**User**: "I need a local AI model for security analysis on a laptop with 8GB RAM"

**Response**: "I recommend the AgentDoG-FG-Qwen3-4B model in Q4_K_M quantization (2.8GB). This provides excellent quality while fitting comfortably in your RAM budget. Download with: `wget https://huggingface.co/mradermacher/AgentDoG-FG-Qwen3-4B-i1-GGUF/resolve/main/AgentDoG-FG-Qwen3-4B.i1-Q4_K_M.gguf` and run with llama.cpp or ollama for privacy-preserving security analysis."

Constraints

Technical Details