mukul975/Anthropic-Cybersecurity-Skills
A comprehensive open-source library featuring 754 structured cybersecurity skills for AI agents, mapped to five industry frameworks to provide expert-level guidance.
Awesome MCP › Security & Reverse Engineering
NVIDIA/SkillSpector
SkillSpector is a security scanner from NVIDIA, written in Python, that inspects AI agent skills and MCP tooling before they are installed. It accepts a Git repository, a URL, a zip archive, a directory or a single file and answers whether the scanned skill is safe to install. Detection covers 68 vulnerability patterns in 17 categories, including prompt injection, data exfiltration, privilege escalation, supply-chain risk, excessive agency, output handling, system prompt leakage, memory poisoning, tool misuse, rogue agent behaviour, anti-refusal, trigger abuse, dangerous code found through AST analysis, taint tracking, YARA signatures and two MCP-specific checks for least privilege and tool poisoning. Analysis runs in two stages: fast static analysis and an optional semantic pass performed by a language model, with the provider configured through environment variables. Supply-chain checks query OSV.dev for current vulnerability data and fall back to offline data when the network is unavailable. Reports are produced for the terminal, as JSON, as Markdown or as SARIF for continuous integration and IDE tooling, and every scan yields a risk score from zero to one hundred with severity labels and recommendations. A baseline mechanism records accepted findings by glob rule or fingerprint so later scans surface only new issues. The tool installs with uv or pip, runs from the included Dockerfile without a local Python setup, and an optional extra adds an MCP mode so scans can be started from inside an agent session. A batch scanner in the contrib directory scans directories of skills in parallel, supports Chinese, Japanese and Korean text and can spread work across several API keys. Ingest limits cap download size and archive member count to bound zip bombs. SkillSpector is part of the NVIDIA Verified Skills pipeline, which scans, evaluates and signs skills before publication.
https://github.com/NVIDIA/SkillSpector
Also in Security & Reverse Engineering
mrexodia/ida-pro-mcp
ida-pro-mcp is an MCP server for IDA Pro that enables advanced reverse engineering capabilities through MCP-based interactions and automation.
LaurieWired/GhidraMCP
GhidraMCP is an MCP server that integrates Ghidra's reverse engineering capabilities with MCP clients, enabling automated binary analysis and decompilation through large language models.
invariantlabs-ai/mcp-scan
MCP-Scan is a security tool that statically and dynamically scans and monitors Model Context Protocol (MCP) connections to detect and prevent vulnerabilities such as prompt injections, tool poisoning, and cross-origin escalations.
zinja-coder/jadx-ai-mcp
JADX-AI-MCP is a JADX plugin integrating Model Context Protocol to enable AI-powered live reverse engineering, vulnerability detection, and code analysis of Android APKs using large language models like Claude.
mariocandela/beelzebub
Beelzebub is a secure low-code honeypot framework leveraging large language models and the Model Context Protocol (MCP) to detect and analyze cyber attacks, including prompt injection attempts against LLM agents.
svnscha/mcp-windbg
mcp-windbg is a Model Context Protocol server that enables AI models to analyze Windows crash dumps using WinDBG through natural language interaction and command execution.
BlackSnufkin/LitterBox
LitterBox is a secure sandbox environment integrating with LLM agents via MCP to provide advanced malware analysis and testing capabilities for red and blue teams before deployment.