ML AI Skills — AI skill for Claude Code

ML / AI Skills Library

 [](.github/workflows/ci.yml) [](LICENSE)

**38 self-contained ML/AI Agent Skills, checked by their own tooling and packaged as a native Claude Code plugin marketplace.**

Overview

Agents are only as good as the procedural knowledge they can load. This repo is a curated library of ML/AI **skills**, one `SKILL.md` per topic, covering everything from data preprocessing to LLM evaluation to AI/ML security. Each skill is either:

• a ⚙️ workflow: a runnable, step-by-step procedure with copy-paste code; or
• a 📖 reference: an explainer an agent loads for context.

It's built for two audiences at once:

• Claude Code users who want reviewed ML/AI guidance available while they work.
• Anyone auditing or extending the library. Every skill follows one formal spec, enforced by a dependency-free Python toolchain (validator, router, eval and benchmark runners, freshness and claims checkers, packaging generators).

It's a content + tooling repo: plain Markdown, YAML frontmatter, generated plugin manifests, a bash installer, and stdlib-only Python. It is not a running service.

**Status: not production-certified.** The skills are guidance to review before use, not verified code. The [Evidence](#evidence-what-has-and-has-not-been-checked) section lists exactly what has been checked, and [`docs/KNOWN-ISSUES.md`](docs/KNOWN-ISSUES.md) lists what has not.

Features

• 38 skills across 8 domains: foundations, classical ML, deep learning, LLMs/GenAI, specialized domains, MLOps, responsible AI, and AI/ML security. Full catalog below.
• Two enforced skill types. Workflow skills require ## Workflow with runnable code; reference skills require ## Key Concepts. Both require ## Gotchas (≥3) and ## References (≥2).