AI Safety Rails

PassAudited by VirusTotal on May 11, 2026.

Findings (1)

The skill is designed to establish safety boundaries and a 'trust ladder' for an AI agent, focusing on prompt injection defense and manual approval workflows. It contains defensive instructions for the agent and installs auxiliary security utilities (ai-sentinel, skill-guard) via npx in SKILL.md. No malicious logic, data exfiltration, or deceptive instructions were found.