The K-12 Lesson Differentiation Agent Skill was co-developed with
Anthropic ↗.
Overview
The K-12 Lesson Differentiation skill adapts an existing K-12 lesson for students at different proficiency levels (below / at / above grade level). It produces one teacher-facing differentiation plan plus three student-ready tier documents in Turn 1, all rendered from one master JSON via bundled scripts. Shared content is written once so tiers cannot drift. The skill uses research-based differentiation principles (Tomlinson framework + subject-specific access design) and works with or without the Learning Commons Knowledge Graph. The skill considers:- Source lesson identification from conversation history, uploads, or URLs
- Subject-specific differentiation rules and document templates
- Curriculum detection (for example, Illustrative Mathematics or OpenSciEd)
- State-specific standards anchoring when jurisdiction signals are present
- Grade-level demand preserved across all tiers
At a glance
Getting started
Follow the Quickstart to start using this skill:Triggers
Your LLM will programmatically load this skill when explicitly or implicitly prompted to differentiate, tier, or scaffold an existing K-12 lesson:- Explicitly: “differentiate this lesson,” “tier for below/at/above,” “scaffold for struggling and advanced students”
- Implicitly: “my students are at different levels,” “I have struggling and advanced students”
Output
Rubrics
The K-12 Lesson Differentiation skill’s outputs are scored by the following rubrics ↗.How to score
- Generate differentiated materials using this skill.
- Apply
differentiation.csvto the tiered artifacts andclarifying_question.csvto evaluate the model’s pre-generation clarification behavior. - Run LLM-as-judge or human review against each criterion. See Evaluating outputs.