A GitHub repo maps 817 security skills to six frameworks

A GitHub repository called Anthropic-Cybersecurity-Skills packages 817 security skills for AI agents. Each skill carries tags against MITRE ATT&CK, NIST CSF 2.0, MITRE ATLAS, D3FEND, NIST AI RMF and MITRE F3, six frameworks security and compliance teams already use to justify budget and staffing decisions. As of August 24, the repo had picked up 2,918 stars.

That mapping is the reason this repo is worth more than a folder of prompts. A skill that says "scan for X" is a convenience, useful but generic. A skill tagged to a specific NIST CSF control is something a security lead can point to when an auditor asks what backs that control.

The repo also packages to the agentskills.io standard, so the same skill set loads across different agent harnesses instead of staying locked to one product.

I use Claude Code every day in my personal projects, and most security-flavored skill packs I've tried are thin, a system prompt and a hope. This one bets that the framework tags do the selling and the skill prompts are secondary. If that holds, expect more of these repos to lead with compliance mapping over prompt count.

The repo doesn't say who's running these skills in production, or whether the framework tags have been checked against real audits. That's the open question before any team wires this into a compliance workflow.