mukul975/Anthropic-Cybersecurity-Skills
This community library packages 817 security workflows as agent skills, offering broad framework-mapped guidance while requiring careful review for offensive and dual-use tasks.
The repository contains **817 skills** across **29 security domains**, using the agentskills.io format. Coverage includes 805 ATT&CK mappings, 804 NIST CSF mappings, and smaller sets for D3FEND, AI RMF, F3, and ATLAS.
Treat it as a searchable playbook layer for security agents: frontmatter supports cheap discovery, while full skill files provide prerequisites, procedures, decision points, and verification steps. Review the selected workflow and required access before execution.
The repository contains **817 skills** across **29 security domains**, using the agentskills.io format. Coverage includes 805 ATT&CK mappings, 804 NIST CSF mappings, and smaller sets for D3FEND, AI RMF, F3, and ATLAS. Treat it as a searchable playbook layer for security agents: frontmatter supports cheap discovery, while full skill files provide prerequisites, procedures, decision points, and verification steps. Review the selected workflow and required access before execution. This is an **independent community project**, not an Anthropic release. It includes offensive and dual-use material, so use must stay within owned systems or explicit written authorization; the repository’s quality claims are not independently validated here.
This expands the skill-catalog pattern into a security-specific playbook layer with unusually broad framework mapping and discovery metadata. Against prior candidates, its scale strengthens the case for selective routing, authorization checks, and supply-chain review rather than wholesale installation. The mappings improve findability, but neither they nor the repository’s own quality claims establish workflow correctness or safe execution.