Llm Integration

This handles local LLM integration through llama.cpp and Ollama, focused on running models securely without cloud dependencies. You get patterns for prompt injection prevention, resource limits to avoid DoS, and model loading with integrity checks. The implementation leans heavily on input sanitization and output filtering since you're processing untrusted prompts. Worth noting it includes specific version requirements tied to CVE fixes (there's a template injection vulnerability in older llama-cpp-python versions). The patterns are built for a voice assistant use case called JARVIS, but the security boundaries and performance optimizations apply to any local LLM deployment where you care about sub-500ms latency and memory constraints.

skills.sh installs: 237
GitHub stars: 48
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add martinholovsky/claude-skills-generator --skill llm-integration --agent claude-code

Installs into .claude/skills of the current project.

Files

Local LLM Integration Skill

File Organization: This skill uses split structure. Main SKILL.md contains core decision-making context. See references/ for detailed implementations.

1. Overview

Risk Level: HIGH - Handles AI model execution, processes untrusted prompts, potential for code execution vulnerabilities

You are an expert in local Large Language Model integration with deep expertise in llama.cpp, Ollama, and Python bindings. Your mastery spans model loading, inference optimization, prompt security, and protection against LLM-specific attack vectors.

You excel at: