anthropics/k12-teacher-skills

Agent Skills for K‑12 Teachers (anthropics/k12-teacher-skills)

What it is – A small open‑source collection of Claude‑agent “skills” that help teachers create and adapt classroom materials. The repo contains the code and an evaluation framework for four pre‑built skills that are shipped with Anthropic’s Claude for Teachers product.

Core skills

Skill	What it does
k12-lesson-plan-creation	Generates a complete, standards‑aligned lesson plan, optionally matching a teacher’s existing curriculum.
k12-lesson-differentiation	Takes an existing lesson and produces tiered versions (below, at, above proficiency) and adaptations for specific student needs while keeping the core content identical.
k12-lesson-prep	Acts as a prep partner: walks a teacher through the key student activity and outputs a concise teacher‑only prep note.
k12-check-for-understanding	Creates a 1‑3 item formative‑assessment check for a math standard, using documented misconceptions as distractors and providing a guide for next‑step instruction.

How it works – The skills are written to call Anthropic’s Learning Commons Knowledge Graph connector, which gives Claude access to U.S. state standards and their progressions. If you run the skills outside of Claude for Teachers, you’ll need to provide a similar knowledge‑graph source or modify the prompts.

Quick start

• Inside Claude for Teachers – No installation; the skills are already available to any Claude for Teachers account.
• Outside Claude – Clone the repo and add it as a plugin:

  git clone https://github.com/anthropics/k12-teacher-skills
  claude plugin marketplace add ./k12-teacher-skills
  claude plugin install k12-education@k12-teacher-skills
  

Repository layout

• plugin/ – The actual skill definitions packaged as a Claude plugin, plus optional third‑party MCP (Model‑Control‑Plane) servers for teachers.
• evals/ – An evaluation harness that lets you run the skills on test prompts and adapt the metrics to your own use case.

Intended audience – K‑12 educators who use Claude for Teachers, as well as developers who want to experiment with AI‑driven lesson‑planning or integrate similar capabilities into other educational tools.

Why it matters – By exposing the prompts and evaluation scripts, the project lets schools and developers inspect, customize, and verify how an LLM generates curriculum‑aligned content, promoting transparency and reproducibility in AI‑assisted teaching.