Skill UI

Protect AI systems from prompt injection attacks by implementing input sanitization, tool permissioning, and output validation. Essential for securing LLM applications, RAG pipelines, and agentic workflows against indirect and direct attacks.

This comprehensive, zero-trust security auditor scans third-party AI agent skills before installation. It performs a rigorous 6-phase review process, analyzing surface patterns, script integrity, permissions, and supply chain risks. It is designed to detect malicious payloads, credential theft, instruction overrides, and prompt injection attacks, ensuring secure AI development and deployment.

Scan agent skills for security issues before adoption. Detects prompt injection, malicious code, excessive permissions, secret exposure, and supply chain risks. Provides static analysis and manual review workflow.

Detects data exfiltration via URL path encoding and chained web_fetch navigation, covering fake trusted UI injection, indirect prompt injection, letter-level path exfiltration, and multi-hop navigation hijacking in agent memory/context scenarios.

Detects prompt injection hidden in legal documents (pleadings, skeletons, evidence) before AI reads them. Scans PDF, Word, HTML, email attachments for invisible text, Unicode smuggling, and AI-directed instructions. Fails closed with optional hook integration.