GarethManning/education-agent-skills

What it solves

This project provides a rigorous, evidence-based foundation for AI in education. It addresses the problem of AI education tools being built on habit and convention rather than empirical research, preventing the scaling of ineffective teaching practices.

How it works

It is a library of 165 pedagogical skills across 20 domains (such as Memory & Learning Science, AI Learning Science, and Inclusive Design). Each skill is a structured prompt with a machine-readable YAML header containing evidence strength ratings, typed input/output schemas, and chaining metadata. These skills can be used manually via copy-paste, installed as plugins for agents like Claude, OpenAI Codex, and Hermes Agent, or accessed programmatically via a Model Context Protocol (MCP) server.

Who it’s for

• Educators: Classroom teachers, university professors, and curriculum designers who need research-grounded support for lesson and assessment design.
• School Leaders: Those in innovative or alternative education contexts (e.g., Montessori, project-based schools).
• AI Builders: EdTech developers and AI researchers who need a structured, programmatically accessible education knowledge layer.

Highlights

• Evidence-Filtered: Only includes skills grounded in named research; explicitly excludes non-empirical frameworks like learning styles.
• Orchestration-Ready: Engineered for programmatic use with YAML schemas and composable outputs rather than simple prompt collections.
• Broad Pedagogical Scope: Covers 20 domains, including both teacher-facing design tools and student-facing live interaction patterns.
• Multi-Agent Support: Compatible with Claude, OpenAI Codex, Hermes Agent, and any tool supporting the Agent Skills standard.