Prompt Engineering

This is built for JARVIS AI assistant work but the patterns translate to any LLM integration where you need to worry about prompt injection. You get layered prompt construction with security guardrails, input sanitization that strips control characters and enforces length limits, and injection detection using regex patterns for common attacks like instruction overrides and delimiter tricks. The task router validates intents against an allowlist before execution, and output validation catches leaked system prompts or tool calls to unauthorized functions. The architecture separates security rules from task instructions from untrusted user input, which is the right approach. Good starting point if you're adding LLM features to an existing system and need defense in depth without overthinking it.

Install to Claude Code

npx -y skills add martinholovsky/claude-skills-generator --skill prompt-engineering --agent claude-code