mthatipamula

Educational Assessment AI Agent Platform — AI skill for Claude Code

AI community

AI agent platform for K-12 assessment: standards-aligned item authoring with RAG and self-review, rubric-based scoring with human review, and a tutoring agent with persistent memory.

How to install Educational Assessment AI Agent Platform

This entry records only its repository, not the path inside it, so there is no exact command to give. Open mthatipamula/educational-assessment-ai-agent-platform and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Educational Assessment AI Agent Platform does

AI agent platform for K-12 assessment: standards-aligned item authoring with RAG and self-review, rubric-based scoring with human review, and a tutoring agent with persistent memory. Safety guardrails for student input, sandboxed code grading, evals and observability. Claude on AWS Bedrock, Python, React, pgvector, MCP.

Alternatives in AI

Nexscope Ecommerce Skills — Ready-to-use AI agent skills for e-commerce intelligence 55 ★
Creative Director Skill — AI creative director with recursive self-assessment: 20+ methodologies (SIT, TRIZ, Bisociation, SCAMPER, Synec 32 ★
Claude Code Prompting 101 — A comprehensive educational repository for mastering prompt engineering with Anthropic's Claude AI, from basic 29 ★

README

Educational Assessment AI Agent Platform

AI agents for K-12 assessment, built to be **reliable and responsible**:

teachers author standards-aligned items with an agent that researches, drafts, reviews and revises its own work;
a scoring agent scores written answers against a rubric, with evidence and confidence, and sends uncertain cases to a teacher;
a tutor agent with persistent memory helps students with hints, never the answer;
student code runs in an isolated sandbox;
every agent is measured by an evaluation framework and visible in an observability view.

**Stack:**

Python 3.12 (FastAPI) with the Anthropic Python SDK, calling Claude on Amazon Bedrock, with the Anthropic API as fallback;
Amazon Titan embeddings on Bedrock, with Voyage AI as fallback;
PostgreSQL + pgvector;
MCP server;
React + TypeScript;
Docker.

Demo data only: synthetic students and items; standard descriptions are paraphrased. No real student data.

Architecture

 Teacher / student (React)  ──▶  nginx :8070  ──▶  FastAPI :8071 ───────────────┐
                                                    │                           │
          ┌─────────────────────────────────────────┼───────────────┐           │ MCP (/mcp): search_standards,
          ▼                     ▼                   ▼               ▼           │ list_items, student_progress
   Authoring agent        Scoring agent        Tutor agent     Safety layer     │
   (Opus: research,       (Sonnet: rubric,     (Sonnet: hints, (regex + Haiku   │
    draft, self-review,    evidence,            tools, memory,  classifier,     │
    revise)                confidence)          leak guard)     PII redaction)  │
          │                     │                   │               │           │
          └────────── Claude gateway: Bedrock ─▶ Anthropic API (fallback) ◀─────┘
                       Embeddings: Titan V2 (Bedrock) ─▶ Voyage voyage-3.5 (fallback)