Anthropic Cybersecurity Skills open-source library released: 818 structured cybersecurity skills mapped to 6 frameworks

The project contains 818 structured cybersecurity skills across 34 security domains. Each skill follows the agentskills.io open standard with YAML frontmatter, a Markdown workflow, and reference files. Skills map to six frameworks by topic: MITRE ATT&CK v19.1 (805 skills), NIST CSF 2.0 (804), MITRE D3FEND (139), NIST AI RMF (97), MITRE F3 (94), and MITRE ATLAS (93). F3 is an anti-financial-fraud framework released in April 2026 that adds the Positioning and Monetization tactics ATT&CK does not cover. Agents first scan the frontmatter of all skills (~30 tokens each), match relevant ones by tag, then load the top three full workflows (500-2000 tokens each), executing and verifying results step by step. This keeps the entire skill-library index within the context window. Developers and security teams using Claude Code, GitHub Copilot, Codex CLI, Cursor, or Gemini CLI can clone the repo or install via npx, letting agents follow senior-analyst workflows for memory forensics, threat hunting, cloud security forensics, and more instead of guessing tool commands. The library is Apache 2.0 licensed, described as a community project unaffiliated with Anthropic PBC.