Anthropic Cybersecurity Skills is a community-created open-source library of structured cybersecurity skills for AI agents. Despite the name, it is independent from Anthropic and packages hundreds of security playbooks across many domains using an agent-skills format that can be loaded into compatible AI platforms.

The library covers defensive security, forensics, cloud and application security, threat analysis, AI security, fraud, and authorized offensive-security techniques. Skills are mapped to relevant industry frameworks including MITRE ATT&CK, NIST CSF, MITRE ATLAS, D3FEND, NIST AI RMF, and the MITRE Fight Fraud Framework, giving the collection a structured taxonomy beyond simple prompts.

The project is designed for lawful security research, defense, education, and authorized testing. It is useful for teams that want agents to follow repeatable security procedures and framework-aligned guidance, but its dual-use material requires appropriate authorization, rules of engagement, and human oversight when applied to real systems.