mark-student-work-multi-agent

What problem does it solve?

This skill provides a robust orchestration layer to automate the marking of a student's completion PDF using a multi-agent architecture. It eliminates manual, error-prone grading workflows by coordinating specialized subagents to map the document structure, transcribe and grade responses, and tag taxonomy, producing a canonical JSON marking artifact and a derived markdown learning report.

Core Features & Use Cases

• Orchestrates a Hierarchical Multi-Agent System that reads a student attempt PDF and, if available, an answer key or teacher annotations, to generate structured marking data.
• Spawns specialized subagents for structural mapping, transcription/grading, and taxonomy tagging, then aggregates their outputs into standardized artifacts.
• Supports both Standard (answer key) and Teacher-Annotated workflows, with clear error handling and retry boundaries; outputs are suitable for auditing, reporting, and learner diagnostics.

Quick Start

Provide the student attempt PDF path and optional answer key, then instruct the orchestrator to run the multi-agent marking workflow to produce a canonical JSON artifact and a learning report.