Two Agent Skills to Help With Prompt Security

A developer shares two agentic skills (llm-prompts:reviewer and llm-prompts:builder) designed to help identify and prevent prompt injection vulnerabilities in LLM applications. The reviewer runs 51 checks drawn from OWASP LLM01:2025, MITRE ATLAS, and NVIDIA NeMo Guardrails against any codebase, rating findings as CRITICAL/HIGH/MEDIUM/LOW with remediation guidance. The builder generates secure system prompts and scaffolding code (Ruby, Python, JS/TS, Go) annotated with check IDs so every defense is documented. The author is transparent that this is not a static analysis tool and results can vary, but argues it raises the security floor for solo developers and small teams building LLM features without dedicated security expertise.