# Anthropic-Cybersecurity-Skills

> A framework-mapped security skill library for coding agents

- Category: Agents & Automation / Agent Skills & Plugins
- Language: Python
- Licence: Apache-2.0
- Repository: https://github.com/mukul975/Anthropic-Cybersecurity-Skills
- Website: https://mahipal.engineer/Anthropic-Cybersecurity-Skills/
- Canonical page: https://ai-tldr.dev/tools/anthropic-cybersecurity-skills/
- Updated: 2026-06-22

## In short

Anthropic-Cybersecurity-Skills is a community-maintained library — by Mahipal Jangra, not by Anthropic — of more than 800 structured cybersecurity skills written for AI agents rather than for people. It sits in the Agent Skills & Plugins category of the Agents & Automation tools and is open-source, written in Python and licensed Apache-2.0.

## Overview

Anthropic-Cybersecurity-Skills is a community-maintained library — by Mahipal Jangra, not by Anthropic — of more than 800 structured cybersecurity skills written for AI agents rather than for people. Each skill is a directory built to the agentskills.io standard: a `SKILL.md` with YAML frontmatter that makes it discoverable, plus optional `references/`, `scripts/` and `assets/` folders holding standards mappings, helper code and templates.

The frontmatter is what makes the collection searchable by an agent: `name`, `description`, `domain`, `subdomain`, `tags` and framework identifiers such as `atlas_techniques`, `d3fend_techniques`, `nist_ai_rmf` and `nist_csf`. Every skill body follows the same four sections — When to Use, Prerequisites, Workflow (step-by-step commands) and Verification — so an agent gets a runnable procedure instead of prose.

The library's distinguishing feature is its framework coverage. Skills are mapped to MITRE ATT&CK v19.1 across 15 tactics, NIST CSF 2.0's six functions, MITRE ATLAS for adversarial ML techniques, MITRE D3FEND for defensive countermeasures, NIST AI RMF 1.0 for AI governance, and MITRE F3 for cyber-enabled financial fraud. Because it targets the agentskills.io standard rather than one vendor's format, it loads without configuration into Claude Code, GitHub Copilot, Cursor, Windsurf, Cline, Aider, Continue, Codex CLI, Gemini CLI and other compatible harnesses. It is Apache-2.0 licensed.

## What it does

- 800+ skills across security domains from cloud and application security to blockchain and fraud
- Built to the agentskills.io standard — YAML frontmatter for discovery, markdown workflow for execution
- Mappings to six frameworks: MITRE ATT&CK, NIST CSF 2.0, MITRE ATLAS, MITRE D3FEND, NIST AI RMF and MITRE F3
- Consistent skill shape — When to Use, Prerequisites, Workflow, Verification — so procedures are runnable, not narrative
- Loads with zero configuration into any agentskills.io-compatible assistant, CLI agent or framework
- Apache-2.0 licensed, so it can be vendored into internal security tooling

## Getting started

Install the skills into a project with the `skills` CLI, or clone the repository if you want to read and adapt them.

1. Add the skills

   The recommended path installs straight from GitHub.

```bash
npx skills add mukul975/Anthropic-Cybersecurity-Skills
```

2. Or clone the repository

   Useful when you want to trim the set or fork skills for an internal standard.

```bash
git clone https://github.com/mukul975/Anthropic-Cybersecurity-Skills.git
cd Anthropic-Cybersecurity-Skills
```

3. Invoke one from your agent

   With the skills installed, ask the agent for the task by name — it matches on the `description` and `tags` in each skill's frontmatter, then follows that skill's Workflow and Verification sections.

```text
Run the threat-modelling skill against this service and map the findings to NIST CSF 2.0
```

## When to use it

- Give a coding agent a repeatable procedure for a security task instead of improvised advice
- Produce findings already mapped to ATT&CK, D3FEND or NIST CSF for a compliance write-up
- Cover AI-specific threats with the ATLAS and NIST AI RMF mappings when reviewing an ML system
- Fork a subset into an internal skill pack that matches your own security standards

## Version history

- 2026-06-22 v1.3.0: v1.3.0 grew the library from 762 to 817 skills, adding AI Security, Supply Chain and Hardware/Firmware domains and MITRE F3 as a sixth framework mapping.