grading-toolkit

grading-toolkit

A Claude Code plugin providing AI-assisted grading skills for academic instructors. Grade any assignment — free response, multiple choice, coding — from PDF submissions to final CSV/JSON export.

Install

/plugin marketplace add <your-github-username>/grading-toolkit
/plugin install grading-toolkit@<your-github-username>
/reload-plugins

Skills

Skill	When to run
/grade-init	Once per assignment — initializes workspace
/grade-profile	Once per instructor/subject — sets grading style
/grade-rubric Q1a	Once per question before grading
/grade-transcribe	Once per assignment — PDF → text
/grade-question Q1a	Per question when ready to grade
/grade-export	On demand — generates CSV/JSON output

Typical Workflow

/grade-init         ← creates workflow.yml, grade stubs, copies scripts
/grade-profile      ← captures instructor style (optional but recommended)
/grade-rubric Q1a   ← calibrate rubric, seed comments.md
/grade-transcribe   ← PDF → transcription files
/grade-question Q1a ← grade all students on Q1a
/grade-question Q1b ← grade all students on Q1b
/grade-export       ← generate grades.csv

Python Scripts

After /grade-init, these scripts live in your project's scripts/ directory:

Script	Usage
check_pdf_type.py <pdf>	Detect text vs. image layer
render_pages.py [project_root]	Batch PDF-to-PNG conversion
extract_text.py [project_root]	Extract text-layer PDFs
check_progress.py [project_root]	Show completion matrix
export_grades.py [project_root]	Generate CSV/JSON output

Requirements

• Python 3.9+ with pdfplumber and pyyaml
• pdftoppm (TeX Live or poppler-utils) for image-layer PDFs
• Claude Code with plugin support