Largest Cybersecurity Skills Library for AI Agents
A GitHub repository known as 'Anthropic Cybersecurity Skills' has been launched as the largest open-source cybersecurity skills library for AI agents. This project, created by the community and not affiliated with Anthropic PBC, contains 762 pre-built skills covering 26 security domains and mapped to six major industry frameworks.
Background: Addressing the Cybersecurity Workforce Gap
According to ISC2, the global cybersecurity workforce gap reached 4.8 million unfilled positions in 2024. AI agents have the potential to help close this gap, but they require structured domain knowledge to function effectively. Although existing agents can write code and search the web, they lack the practitioner guides needed to act like senior security analysts. This project fills that gap by providing an AI-native knowledge base built from the ground up for standard agentskills.io.
Content: 762 Skills Across 26 Domains
Each skill follows a consistent directory structure, including a SKILL.md file with YAML frontmatter and Markdown body, as well as reference directories, scripts, and assets. The covered domains include:
- Cloud Security (60 skills)
- Threat Hunting (55)
- Threat Intelligence (50)
- Web Application Security (42)
- Network Security (40)
- Malware Analysis (39)
- Digital Forensics (37)
- Security Operations (36)
- Identity and Access Management (35)
- SOC Operations (33)
- Container Security (30)
- OT/ICS Security (28)
- API Security (28)
- Vulnerability Management (25)
- Incident Response (25)
- Red Team (24)
- Penetration Testing (23)
- Endpoint Security (17)
- DevSecOps (17)
- Phishing Defense (16)
- Cryptography (14)
- Zero Trust Architecture (13)
- Mobile Security (12)
- Ransomware Defense (7)
- Compliance and Governance (5)
- Fraud Technology (2)
Mapping to Six Industry Frameworks
The uniqueness of this project is mapping each skill to six frameworks simultaneously:
- MITRE ATT&CK v19.1: 15 tactics, 286 techniques - covering 754 skills
- NIST CSF 2.0: 6 functions, 22 categories - whole organization
- MITRE ATLAS v5.4: 16 tactics, 84 techniques - AI/ML threats
- MITRE D3FEND v1.3: 7 categories, 267 techniques - defense steps
- NIST AI RMF 1.0: 4 functions, 72 subcategories - AI risk management
- MITRE Fight Fraud Framework (F3) v1.1: 8 tactics, 123 techniques - cyber financial fraud
For example, the skill of analyzing malware network traffic is mapped to T1071 (ATT&CK), DE.CM (NIST CSF), AML.T0047 (ATLAS), D3-NTA (D3FEND), and MEASURE-2.6 (AI RMF).
How AI Agents Use These Skills
Each skill requires about 30 tokens to scan the frontmatter and 500โ2,000 tokens to load the entire workflow. This allows agents to search all 762 skills in one pass without overflowing the context window. For instance, when a user asks "Analyze this memory dump for signs of credential theft," the agent will:
Supported Platforms
This project is compatible with more than 20 platforms including Claude Code, GitHub Copilot, Cursor, Windsurf, OpenAI Codex CLI, Gemini CLI, Devin, Replit Agent, LangChain, CrewAI, and others that support the agentskills.io standard.
Status and Community
This repository has received 18,300 stars and 2,200 forks on GitHub. The latest version v1.2.0 (April 5, 2026) features coverage of five frameworks. The project welcomes community contributions, especially for underdeveloped domains such as Fraud Technology and Compliance. Each PR is reviewed within 48 hours.
Conclusion
'Anthropic Cybersecurity Skills' offers a structured knowledge base that enables AI agents to act like senior security analysts. With comprehensive mapping to industry frameworks and a progressive design, this project has the potential to accelerate AI use in cybersecurity operations and help address the global workforce gap.
---
*Original source: [github.com](https://github.com/mukul975/Anthropic-Cybersecurity-Skills)*