Skills that bring expert pedagogy into AI systems
Open and reusable to improve how AI agents complete instructional tasks
Extend model behavior with guidance built for real teaching and learning tasks.
Improve the reliability of educational outputs
Turn instructional best practices into guidance that AI agents can apply
General-purpose models don’t always know how to apply the instructional practices that reflect the best teaching and learning in a classroom. Agent Skills codify expert knowledge, learning science, and pedagogy into guidance AI agents can use when completing education-specific tasks.
Use them to build experiences for lesson plan creation, differentiation, lesson prep, or checking for student understanding, as well as other instructional needs.
For example, if you ask an AI model to differentiate a lesson, it might simply make an assignment easier. But our Lesson Differentiation Skill instead helps guide the model to preserve grade-level rigor while adding the scaffolds, supports, and teacher guidance students need to access the same core learning goal.
Built through a rigorous test and iteration process
Start with expertise and evidence
We identify high-value educational tasks, study how models perform on them today, and work with learning scientists, educators, and edtech experts to define a stronger approach.
Define what good looks like
Each skill is developed alongside an open evaluation rubric that makes the quality bar explicit, from pedagogy and rigor to usability and model behavior. The rubric becomes the standard we test against.
Test, measure, and refine
We run skills through a repeatable evaluation harness to compare outputs against that quality bar at scale. We use the results to identify where a skill falls short, refine the guidance, and test again.
Guide your agents on how to approach education-specific work
Expand product capabilities
Start with a research-backed approach to a key instructional problem, so you aren’t starting agents from scratch and can focus on shaping the experience for your users.
Integrate expert pedagogy
Use our Agent Skills as guidance for a baseline, then layer in the pedagogical choices and product decisions that make your product experience distinct.
Adapt what works
Start with our Agent Skills as a reference point. Inspect the guidance. Compare it with your own approach. Test what works for your product. Fork it, extend it, or incorporate useful elements.
Refer to the high-quality evaluation
Each Agent Skill is built and tested against open evaluation rubrics that make educational quality transparent and measurable. By making the quality bar visible, you can inspect how we evaluate a skill and use the same criteria to assess and improve your AI experiences.
