The Grading & Feedback Prompt Workshop

A configurable tool for building consistent, rubric-aligned AI grading prompts. Fill the fields, upload your key or rubric, generate a ready-to-paste prompt.

A tool for teachers · Designed for Claude, ChatGPT, or Gemini · Inline HTML, no server required

Assignment Context

Tell the AI what it's grading and who it's grading for.

Course

Grade Level

Assignment Title

Assignment Type

Learning Objectives

The AI will reference these when scoring. Be specific — "Simplify radical expressions using prime factorization" beats "work with radicals."

II.

Rubric & Point Values

Either upload your answer key or rubric file, or build the rubric here directly. You can do both.

Total Points Possible

Upload Answer Key or Rubric (Optional)

↥

Drop a file here, or click to browse

PDF, DOCX, TXT, or image · Max 10 MB

The file won't be uploaded anywhere — it's just attached when you paste the prompt into Claude/ChatGPT. You'll attach it again in the chat window.

Rubric Criteria (Optional if Uploading a File)

Partial Credit Rules

III.

Feedback Depth & Granularity

Two independent choices: how much detail, and whether it's delivered overall or item-by-item.

— Depth —

Brief

Shortest usable comment. One or two sentences naming the most important fix. Fast turnaround for low-stakes work.

Moderate

A strength, 2–3 specific errors with corrections, and a next-step suggestion. Good default for quizzes and problem sets.

Detailed

Full breakdown with strengths, all errors with corrections, targeted practice, and growth summary. Best for major assessments, essays, and labs.

— Granularity —

Overall feedback

One consolidated set of comments about the whole submission. Best when you want a holistic view and the depth of feedback refers to the assignment as a whole.

Per-item feedback · question-by-question

A comment for each question or problem. The depth chosen above applies to each item. Best for problem sets, quizzes, and tests where students need targeted feedback on specific questions.

Tip: "Brief + Per-item" works well for math and physics problem sets — one short confirmation or correction per question. "Detailed + Overall" is ideal for essays. "Moderate + Per-item" is the richest option for quizzes with 5–15 items.

IV.

Output Format

Where will the feedback end up?

Canvas-ready plain text

Paste directly into Canvas SpeedGrader comment box. No special characters or formatting.

Canvas-ready HTML · Xavier-branded

Inline CSS with Xavier's black and gold (#827554) palette, Georgia headers, Arial body. For Canvas pages or styled comments.

Plain Markdown

Universal format. Copy-paste into Word, Google Docs, or any document processor.

PDF feedback sheet · matches student filename

One-page PDF with Xavier letterhead. Output filename matches the student's submission (e.g., Smith_J_Quiz4.pdf → Smith_J_Quiz4_feedback.pdf). Good for returning graded work by email or upload.

Structured JSON

Machine-readable fields (student_name, total_score, criterion_scores, strengths, errors, next_steps). For gradebook imports or batch processing.

Voice & Additional Notes

Optional — add context the AI should keep in mind.

Special Instructions (Optional)

Ready? Your configured prompt will appear below.

Your Grading Prompt

Paste this into Claude, ChatGPT, or Gemini. If you uploaded a rubric file above, remember to also attach it to your chat conversation — this tool doesn't send files, it just builds the instruction text.