Connect

SkillsSafe

GUCCI-atlasvbyGUCCI-atlasv
•
API Development
Developer Tools
Security & Testing

Scans AI agent skill files, Model Context Protocol configurations, and system prompts to detect security threats like credential theft, data exfiltration, and shell injection.

As AI agents become more sophisticated, they are increasingly vulnerable to malicious skills that can lead to credential theft, data exfiltration, or agent hijacking. SkillsSafe acts as a critical pre-installation security layer, akin to an antivirus for software, by comprehensively scanning SKILL.md files, MCP configurations, and system prompts. It identifies and mitigates various threats including hidden prompt injections, zero-width character attacks, and attempts to execute arbitrary commands, providing a risk score and detailed reports to ensure the safety and integrity of AI agent operations.

Key Features

01REST API for integration with any agent, script, or CI/CD pipeline.
020 GitHub stars
03Zero-Width Character Detector to visualize invisible Unicode characters embedded in text.
04Detects credential theft, data exfiltration, prompt injection, shell injection, and zero-width character threats.
05Comprehensive Security Scanner for AI skills, MCP configs, and system prompts.
06Native MCP Server Integration allowing agents to automatically check skill safety.

Use Cases

01Integrating security scanning into CI/CD pipelines for AI agent development.
02Pre-installation scanning of AI agent skills and system prompts for security vulnerabilities.
03Automated safety checks for AI skills via Model Context Protocol (MCP) integrations.