finds.dev

mukul975/Anthropic-Cybersecurity-Skills

754 structured cybersecurity skills for AI agents · Mapped to 5 frameworks: MITRE ATT&CK, NIST CSF 2.0, MITRE ATLAS, D3FEND & NIST AI RMF · agentskills.io standard · Works with Claude Code, GitHub Copilot, Codex CLI, Cursor, Gemini CLI & 20+ platforms · 26 security domains · Apache 2.0

View on GitHubHomepage ↗

★ 12,623PythonApache-2.0updated May 2026

// the take

A structured library of 754 cybersecurity skill definitions in YAML+Markdown format, mapped to MITRE ATT&CK, NIST CSF 2.0, ATLAS, D3FEND, and NIST AI RMF. Aimed at teams wanting to give AI coding assistants or autonomous agents a consistent playbook for security tasks rather than relying on the model's baked-in knowledge. Think of it as a prompt/context library with framework crosswalks, not executable tooling.

- The agentskills.io format with lightweight YAML frontmatter (~30 tokens) plus full Markdown body is a smart progressive-disclosure design — agents can scan all 754 skills cheaply and only load full context for matched ones.

- Five-framework crosswalk (ATT&CK + NIST CSF + ATLAS + D3FEND + AI RMF) in one place is genuinely rare; finding which D3FEND countermeasure maps to which ATT&CK technique normally requires manual work across three separate sites.

- Coverage breadth is real — OT/ICS, container security, mobile, API security, and cloud forensics all have dedicated skills, not just the usual web app + network basics that most security repos stop at.

- Apache 2.0 license, CI validation workflow, and ATT&CK Navigator export layer are practical additions that make this actually usable in enterprise settings, not just a README demo.

- The name 'Anthropic Cybersecurity Skills' is misleading — this has nothing to do with Anthropic PBC. It's a community project by a different author that hijacks the brand name for star farming. The disclaimer is buried at the bottom of a very long README.

- Skill quality is uneven and hard to audit at scale. Many skills have thin reference files and the agent.py scripts appear to be boilerplate rather than tested, working code. The 4x star explosion in a short period (visible in the star history shape) suggests viral promotion, not organic adoption by practitioners.

- No test harness or benchmark exists to verify whether following a skill actually produces better agent outcomes than not having it — the value proposition is entirely theoretical and unvalidated.

- The 'agentskills.io standard' it references is the author's own website, not a recognized community standard, making the 'compatible with 20+ platforms' claim marketing rather than fact — any platform that reads Markdown files would 'support' this.

We read the README, the directory tree and a key source file in May 2026. Nobody pays to be listed. How we judge →

// at a glance

Stars
12,623
Forks
1,463
Language
Python
License
Apache-2.0
Last commit
30 May 2026

// nearby

If Anthropic-Cybersecurity-Skills isn't quite it.

All repos →