This repository offers a library of 152 evidence-based pedagogical skills for AI agents, addressing the need for rigorous, research-grounded educational practices in AI tools. Targeting educators, AI developers, and researchers, it aims to enhance teaching quality and enable personalized learning by leveraging validated educational science over conventional methods.

How It Works

Pedagogical knowledge is structured into discrete, AI-consumable skills, each meticulously grounded in named research and rated for evidence strength. The library explicitly excludes unproven frameworks like learning styles, documenting these exclusions. Skills are engineered for programmatic use with YAML schemas, typed inputs/outputs, and chaining metadata, facilitating robust AI agent orchestration.

Quick Start & Requirements

Installation varies: Claude users add the plugin via its GitHub URL. Codex users install it locally. General Agent Skills compatibility involves copying skill folders. Manual use requires no setup, only copying prompt blocks. Hosted MCP access via Vercel requires an auth token; local installation is recommended for free, sustainable use.

Highlighted Details

• 152 skills across 19 domains, covering memory, self-regulation, curriculum design, and AI learning science.
• Explicit documentation of excluded, unevidenced frameworks (e.g., learning styles) in EXCLUSIONS.md.
• Transparent evidence strength ratings (Strong, Moderate, Emerging, Original) for each skill.
• Skills designed for AI orchestration with machine-readable YAML headers, typed inputs/outputs, and chaining metadata.

Maintenance & Community

Maintained by educator Gareth Manning (20 years international experience). Contributions via GitHub Issues/PRs. Contact: gareth.manning@gmail.com, X: @worldteacherman.

Licensing & Compatibility

Licensed under CC BY-SA 4.0. Permits open use, forking, and sharing with attribution and under the same license. Compatible with open-source and commercial applications, subject to the ShareAlike clause.

Limitations & Caveats

Hosted MCP server access now requires an authentication token. Skill prompts lack empirical validation as AI interventions, though underlying research is validated. Skill changes require rebuilding and committing the MCP server bundle for live server updates.