Add VSOS Guard — zero-dependency prompt injection defense for AI agents

[woshilaohei]

Project Submission

Name: VSOS Guard
URL: https://github.com/woshilaohei/vsos-guard
PyPI: https://pypi.org/project/vsos-guard/

Description

Zero-dependency runtime input guard for AI agents. Pure rule engine that intercepts prompt injection, jailbreak attempts, and adversarial instructions before they reach the LLM. No API keys, no external services, sub-5ms latency.

Why it fits awesome-mcp-security

• Input boundary defense: Catches adversarial inputs at the agent boundary — before they enter the context window
• Zero dependencies: Pure Python rule engine, no LLM-in-the-loop for safety checks
• MCP-compatible: Can be used as a pre-processing layer before any MCP tool invocation
• OWASP LLM Top 10 coverage: Addresses LLM01 (prompt injection), LLM06 (sensitive data disclosure), LLM07 (system prompt leakage)
• Open source: MIT license, pip install vsos-guard

Category

Defense / Guardrails