HoneyBadger

•
4
•

Scans AI agent skills and MCP servers for secrets, CVEs, supply chain attacks, and prompt injection before installation.

HoneyBadger acts as a critical pre-installation security gate for AI agent skills and MCP servers, performing comprehensive static analysis without executing code. It meticulously scans for a wide array of threats including secrets, CVEs, supply chain attacks, and prompt injection vulnerabilities within SKILL.md files. Engineered as a single, cross-platform Go binary with an offline mode and configurable paranoia levels, it offers robust protection for various AI runtimes like Claude Code, OpenClaw, and NanoBot, while seamlessly integrating into CI/CD pipelines.

주요 기능

01Comprehensive static analysis for AI agent skills (secrets, CVEs, supply chain, prompt injection)
02Configurable paranoia levels for tailored security enforcement
03Operates as a single, cross-platform Go binary with full offline capability
04Native MCP server mode via JSON-RPC for agent runtime integration
05Supports pre-install gating for various AI platforms and CI/CD pipelines
060 GitHub stars

사용 사례

01Enforcing security policies as a pre-installation gate for AI agent skills
02Integrating automated security scanning into CI/CD pipelines
03Performing on-demand security audits of AI agent codebases and skill manifests