CLI Abliterated

Local GGUF-powered CLI assistant for security researchers.

Features

• Local Inference: Run models entirely offline using GGUF format
• Interactive Chat: Terminal-based conversation interface
• File Context: Add files/directories with @path syntax
• Slash Commands: Built-in commands for session management
• Conversation History: Persistent history across sessions
• 100% Offline: No network calls, complete privacy
• Security-Focused: Built-in pentest tool integration

Default Model

Qwen2.5-Coder-3B-Instruct-Distill-Qwen3-Coder-Next-abliterated (Q4_K_M)

• 3B parameters, fast inference
• 4-bit quantization for efficiency
• Distilled from larger Qwen3-Coder-Next

Security Commands

Command	Description
/scan:analyze <file>	Analyze nmap, sqlmap, nikto scan results
/report:generate	Generate pentest report sections
/cve:lookup <CVE-ID>	Look up CVE information
/payload:generate	Generate payload suggestions
/findings:prioritize	Prioritize security findings

Installation

Quick Install

cd /home/gnul/cli-abliterated

# Create virtual environment
python3 -m venv .venv
source .venv/bin/activate

# Install dependencies
pip install -e .

# Run
cli-abliterated -m /path/to/model.gguf

Dependencies

• Python 3.10+
• llama-cpp-python
• rich
• prompt-toolkit

Usage

Interactive Mode

# Auto-detect model
cli-abliterated

# Specify model
cli-abliterated -m /path/to/model.gguf

# With context
cli-abliterated -c ./src -c ./docs

Single Prompt

cli-abliterated -p "Explain this code" -c ./main.py

Commands

Standard Commands

Command	Description
/help	Show available commands
/clear	Clear conversation history
/model	Show model information
/stats	Show session statistics
/context	Show loaded context files
/save [file]	Save conversation to file
/exit	Exit the application

Security Commands

Command	Description
/scan:analyze <file>	Parse nmap, sqlmap, nikto, Burp, Nuclei results
/vuln:research <type>	Research vulnerabilities (OWASP, CWE, CVSS, remediation)
/cvss:calculate	Calculate CVSS 3.1 score from metrics
/report:generate	Generate pentest report sections
/cve:lookup <CVE-ID>	Look up CVE information
/payload:generate	Generate reverse shell payloads
/findings:prioritize	Prioritize findings by severity
/evidence:init <name>	Initialize evidence directory structure
/surface:map	Generate attack surface visualization

File Context

Use @ to add files or directories to context:

>>> @./src/main.py Explain this function

>>> @./src/ Add authentication to this codebase

>>> @~/.bashrc Help me optimize this config

Security Workflow Examples

# Analyze nmap scan results
/scan:analyze /path/to/scan.xml

# Analyze sqlmap output  
/scan:analyze /path/to/sqlmap_output.txt

# Analyze Burp Suite export
/scan:analyze /path/to/burp_export.xml

# Analyze Nuclei results
/scan:analyze /path/to/nuclei_results.json

# Research a vulnerability
/vuln:research SQL Injection
/vuln:research XSS
/vuln:research SSRF

# Calculate CVSS score
/cvss:calculate AV:network AC:low PR:none UI:none S:unchanged C:high I:high A:high

# Look up CVE details
/cve:lookup CVE-2024-1234

# Initialize evidence directory
/evidence:init client_engagement_2024

# Generate report section
@scan_results.xml Generate executive summary for this scan

# Code review for security issues
@./modules/exploit.py Review this exploit code for bugs

Configuration

Option	Description	Default
-m, --model	Path to GGUF model	Auto-detect
-c, --context	Add context paths	None
-n, --max-tokens	Max tokens to generate	1024
-t, --temperature	Temperature (0-2)	0.7
--n-gpu	GPU layers (-1=all)	-1
--n-ctx	Context window size	4096

Project Structure

cli-abliterated/
├── src/qwen_local/
│   ├── __init__.py
│   ├── cli.py        # CLI entry point
│   ├── model.py      # Model manager
│   ├── chat.py       # Chat session
│   ├── context.py    # Context manager
│   └── security.py   # Security commands (nmap, sqlmap, CVE)
├── cli-abliterated.sh     # Launcher script
├── install.sh        # Installation script
├── pyproject.toml
└── README.md

License

MIT