微步旗下AI智能体Skills安全平台
Secure EveryAI Agent Skill
No file selected
🔒 Files are analyzed securely⚡ Results in <30 seconds🚀 Start for free
Core Capabilities
Two Pillars of Agent Skill Security
Comprehensive threat detection combined with a curated marketplace of verified safe skills.
Skill Scan
Multi-layered security analysis engine for AI Agent Skills. Detect prompt injection, data exfiltration, malicious code, and supply chain threats.
AgentScanner static analysis with Yara/YAML rules & AST
LLM-powered intent analysis and anomaly detection
URL deep inspection with threat intelligence correlation
Sub-file detection: scripts, binaries, markdown files
OWASP ASI Top 10 coverage
Skill Hub
A curated marketplace of security-verified AI Agent Skills. Every skill is scanned, scored, and badged before listing.
Security-verified skill packages with trust scores
Categorized browsing across 8+ domains
Developer profiles with reputation tracking
One-click install for Cursor, Claude, VS Code & more
Community reviews and vulnerability reporting
Detection Engine
4-Layer Detection Pipeline
From input validation to comprehensive verdict, every skill goes through rigorous multi-dimensional analysis.
Input Layer
Accept skill packages via upload, URL, or name search. Auto-download from ClawHub and other registries.
Static Analysis
AgentScanner performs metadata extraction, pattern matching with Yara/YAML rules, AST analysis, and local SLM intent detection.
Deep Analysis
LLM intent analysis, URL threat intelligence correlation, sub-file deep inspection with sandbox execution.
Verdict
Multi-dimensional weighted scoring and rule-based judgment. Generate structured report: Safe / Suspicious / Malicious.