SecureShell v0.2.0 - Plug-and-Play Security for LLM Agents

SecureShell v0.2.0

SecureShell is a plug-and-play security layer that acts as "sudo for LLMs" - preventing AI agents from executing dangerous shell commands.

What is SecureShell?

LLM agents with shell access can hallucinate destructive commands like rm -rf / or dd if=/dev/zero. SecureShell solves this with:

• Zero-Trust Gatekeeper - Every command validated by independent LLM before execution
• Platform-Aware - Blocks Unix commands on Windows (and vice versa)
• 3-Line Integration - Drop into existing LangChain, LangGraph, or MCP projects
• Multi-LLM Support - Works with OpenAI, Claude, Gemini, Groq, Ollama & more

Key Features

Security First

• Risk Classification: Automatic GREEN/YELLOW/RED categorization
• Sandbox Validation: Path-based access control
• Audit Logging: Track every command execution
• Agent Feedback: Clear explanations help LLMs learn from mistakes

Plug-and-Play Integration

# Python - 3 lines to secure your agent
from secureshell import SecureShell
shell = SecureShell(template='strict')
result = await shell.execute('ls -la', 'List files')