Score a piece of work against a written rubric, one criterion at a time, quote the evidence for each mark, and say what would reach the top level, using only the rubric you give.
Sample results: runs and scores on this page are illustrations while testing is set up. How we test
Who it is for: Anyone who marks or reviews written work against a rubric, such as a committee member, a manager, a teacher or a program lead.
Select a part to see its prompt, input, output and check.
The same job, with and without the template
Sample
One plain prompt
Score this proposal against my rubric, one criterion at a time, using the marks in the rubric, and say what would improve it. (The same rubric and proposal are pasted below.)
Overall this is a solid proposal that meets most of the rubric.
Clear aim: met. It names small shop owners and says the workshop will help them see which products make money.
Realistic plan: met. Four two-hour sessions led by a volunteer accountant is realistic.
Fits the audience: met. A public library is a central place for shop owners, and weekday mornings are a sensible time for a workshop.
Cost: met. The proposal asks for 1,200 dollars and says the workshop is free.
Way to tell if it worked: partly met. Owners will bring their own receipts, so the leader will be able to see their progress.
To improve, consider adding a short survey at the end.
Problem:It marks the cost met with no breakdown, calls weekday mornings a fit for owners who are open from 9 am to 6 pm, and credits a way to tell if it worked that the proposal never gives.
"Most open at 9 am and close at 6 pm ... on weekday mornings at the public library"
Move the sessions to a day or time that suits owners who are open from 9 am to 6 pm.
4
"Cost"
partly met
"We will charge nothing, and we ask for 1,200 dollars"
Show what the 1,200 dollars buys, as well as the total.
5
"Way to tell if it worked"
not met
No passage in the work
Name something that will be counted or asked at the end of the workshop.
Met
"Clear aim"
"Realistic plan"
Check:Every description of met is quoted word for word from the rubric: Five criteria, each description of met quoted word for word from the rubric.
Check:Every passage is quoted word for word from the work: Five rows: four passages quoted word for word from the work, and None found for the way to tell if it worked.
Check:Each mark is one of the three allowed words: Five marks, each one of the three allowed words: 2 met, 2 partly met, 1 not met.
Check:Every row to improve is filled in: Three rows under To improve, each with all five cells filled.
Check:Every criterion in the report is in the rubric word for word: All three quoted criteria are in the rubric word for word.
Check:The counts add up to the marks: The Result line says 5 criteria: 2 met, 2 partly met and 1 not met; the table has three rows, one for each criterion not marked met.
Check:Each mark and each fix follows the rubric's own words: Each mark matches the rubric's wording for its level, and each To reach met uses only the rubric's words and facts already in the work (9 am, 6 pm and 1,200 dollars).
For your read:You decide whether you agree with each mark: Do you agree that weekday mornings do not suit shop owners who open at 9 am?
Runs and scores
Sample
Claude, Sonnet model
October 4, 2026, easy sample
Check:Every description of met is quoted word for word from the rubric
Two criteria, each description of met quoted word for word from the rubric.
Check:Every passage is quoted word for word from the work
Two passages, each quoted word for word from the work.
Check:Each mark is one of the three allowed words
Two marks, both met.
Check:Every row to improve is filled in
One row under To improve, with None found in every cell, because both criteria are met.
Check:Every criterion in the report is in the rubric word for word
The report quotes no criterion under To improve (its row says None found), so every quote in it is in the rubric.
Check:The counts add up to the marks
The Result line says 2 criteria marked, both met; the table has no row for a criterion.
Check:Each mark and each fix follows the rubric's own words
Both marks match the rubric's wording for met: the note names the soup run and the 60 households.
For your read:You decide whether you agree with each mark
The note says Because of you, 60 households got a hot meal, so do you agree that it describes one result?
Fixed: Nothing needed fixing: every check passed on the first pass.
Met 8 of 8
1 of these is for your read: the tool hands it to you.
ChatGPT, GPT model
October 5, 2026, tricky sample
Check:Every description of met is quoted word for word from the rubric
Five criteria, each description of met quoted word for word from the rubric.
First pass failed:Every passage is quoted word for word from the work
The first version gave the passage for the quote criterion as "a customer said orders are never wrong now", which is not in the work word for word.
Check:Each mark is one of the three allowed words
Five marks, each one of the three allowed words: 2 met and 3 partly met.
Check:Every row to improve is filled in
Three rows under To improve, each with all five cells filled.
Check:Every criterion in the report is in the rubric word for word
All three quoted criteria are in the rubric word for word.
Check:The counts add up to the marks
The Result line says 5 criteria marked: 2 met, 3 partly met and 0 not met; the table has three rows, one for each criterion not marked met.
Failed:Each mark and each fix follows the rubric's own words
The To reach met for the next-step row suggests adding a link to the order form, which neither the rubric nor the work gives.
For your read:You decide whether you agree with each mark
The five hours a week has no source beside it, so will you ask the owner where it comes from before you use it?
Fixed: The evidence step ran again once with the rule to quote word for word, and it came back quoting the whole sentence about the regular customer, which cleared the first failure. The report step ran again once and the next-step row still suggested a link to the order form, so the check failed again and the run stopped there and flagged it for you.
Met 7 of 8
Not met:Each mark and each fix follows the rubric's own words: The next-step fix still suggests a link to the order form, which nobody gave.
1 of these is for your read: the tool hands it to you.
n8n, import format 2.20
October 6, 2026, typical sample
Check:Every description of met is quoted word for word from the rubric
Five criteria, each description of met quoted word for word from the rubric.
Check:Every passage is quoted word for word from the work
Five rows: four passages quoted word for word from the work, and one None found.
First pass failed:Each mark is one of the three allowed words
Row 3 said "mostly met", and the Mark column must say met, partly met or not met.
Check:Every row to improve is filled in
Three rows under To improve, each with all five cells filled.
Check:Every criterion in the report is in the rubric word for word
All three quoted criteria are in the rubric word for word.
Check:The counts add up to the marks
The Result line says 5 criteria marked: 2 met, 2 partly met and 1 not met; the table has three rows.
Check:Each mark and each fix follows the rubric's own words
Each mark matches the rubric's wording for its level, and each To reach met uses only the rubric's words and facts already in the work.
For your read:You decide whether you agree with each mark
Will you decide whether weekday mornings really do not suit the shop owners before you pass on the marks?
Fixed: The marks step ran again once with the three allowed words restated, and every row then used one of them.
Met 8 of 8
1 of these is for your read: the tool hands it to you.
The steps, and the check on each
1
Step 1: List each criterion and what met looks like
Split the rubric into separate criteria, quoted word for word, with the rubric's own words for the top level.
Hands back: A numbered table of the criteria, each with the rubric's own words for met.
Check:Every description of met is quoted word for word from the rubric
A plain rule. If it fails, run the step again (the n8n file does this for you).
2
Step 2: Find the passage for each criterion
For each criterion, quote the words in the work that bear on it, or say that none do.
Hands back: A table that puts a quoted passage from the work beside each criterion, or None found.
Check:Every passage is quoted word for word from the work
A plain rule. If it fails, run the step again (the n8n file does this for you).
3
Step 3: Mark each criterion
Compare each passage with the rubric's wording and give one of three marks, with the reason in one sentence.
Hands back: A table that gives each criterion one of three marks and the reason.
Check:Each mark is one of the three allowed words
A plain rule. If it fails, run the step again (the n8n file does this for you).
4
Step 4: Write the marking report
Turn the marks into a count, a list of what to improve with wording taken from the rubric, and a list of what is met.
Hands back: A report with a count line, a table of what to improve and a list of the criteria that are met.
Check:Every row to improve is filled in
A plain rule. If it fails, run the step again (the n8n file does this for you).
After the last step
Check:Every criterion in the report is in the rubric word for word
A plain rule. If it fails, run the step again (the n8n file does this for you).
Check:The counts add up to the marks
A second look by the AI. If it fails, run the step again (the n8n file does this for you).
Check:Each mark and each fix follows the rubric's own words
A second look by the AI. If it fails, run the step again (the n8n file does this for you).
Check:You decide whether you agree with each mark
You check this one. If it fails, it is flagged for you.
What you need first
The rubric(needed)
Paste the rubric as written, with what each level looks like for each criterion. The marks used here are met, partly met and not met.
The work to mark(needed)
Paste the full text of the work as it stands. Remove names you do not want an AI tool to see first.
Ways to adapt it
This template marks by the rubric's own words and nothing else, so a vague rubric gives vague marks. If a criterion says only "good" or "clear", add what each level looks like before you run it.
If your rubric uses other level names, such as strong, adequate and weak, change the three words in the marks step and in its check, both.
If your rubric gives points, keep them out of the AI's job. Let it mark in words, then add the points yourself from the marks.
To check a list of must-haves rather than quality at several levels, use Check every requirement is covered instead.
For a long piece of work, run it on one section at a time, with only the criteria that apply to that section.
Change any step or check in the Editor, then download it again.
A short proposal marked on five criteria; two are met, two are partly met (a time that does not suit the audience, a cost with no breakdown) and one is not met.
The rubric
Rubric for workshop proposals. Each criterion is marked met, partly met or not met.
1. Clear aim. Met: the proposal states who the workshop is for and what they will be able to do afterward. Partly met: it names the audience or the outcome but not both. Not met: it names neither.
2. Realistic plan. Met: it gives the number of sessions, the length of each and who will lead them. Partly met: it gives two of the three. Not met: it gives one or none.
3. Fits the audience. Met: the day, time and place suit the audience the proposal names. Partly met: the topic suits them but the day, time or place does not, or the reverse. Not met: the proposal names no audience, or nothing suits them.
4. Cost. Met: it gives a total cost and shows what the money buys. Partly met: it gives a total with no breakdown. Not met: it gives no cost.
5. Way to tell if it worked. Met: it names something that will be counted or asked at the end. Partly met: it says it will gather feedback without saying how. Not met: it says nothing about this.
The work to mark
Proposal: Spreadsheet Skills Workshop. Many small shop owners in our town cannot say which products make money. Most open at 9 am and close at 6 pm. We will help them get confident enough with spreadsheets to find out. The workshop runs for four sessions, two hours each, on weekday mornings at the public library. Our volunteer accountant, Ravi Mehta, will lead every session, and owners will bring their own receipts. We will charge nothing, and we ask for 1,200 dollars.
The scoring guide
Sample
One criterion for each check, marked met or not met. (The guides in Quality rate a whole piece of work on three levels.) Under each one is how the three sample runs scored.
Every description of met is quoted word for word from the rubric
A plain rule
Meets:Every description of met in the table is found word for word in the rubric.
Does not meet:A description has been reworded or improved, or is not in the rubric word for word.
Claude:Met
ChatGPT:Met
n8n:Met
Every passage is quoted word for word from the work
A plain rule
Meets:Every passage is found word for word in the work (a quote joined with ... is checked piece by piece), or the row says None found.
Does not meet:A passage has been reworded, or is not found word for word in the work.
Claude:Met
ChatGPT:Met
n8n:Met
Each mark is one of the three allowed words
A plain rule
Meets:Every criterion has exactly one of met, partly met or not met.
Does not meet:A row has another word, such as mostly met or good.
Claude:Met
ChatGPT:Met
n8n:Met
Every row to improve is filled in
A plain rule
Meets:Every row under To improve has the criterion, the mark, the evidence and what would reach met filled in.
Does not meet:A row under To improve has a blank or a dash in one of its cells.
Claude:Met
ChatGPT:Met
n8n:Met
Every criterion in the report is in the rubric word for word
A plain rule
Meets:Every criterion quoted in the report is found word for word in the rubric.
Does not meet:A quoted criterion has been reworded or is not in the rubric.
Claude:Met
ChatGPT:Met
n8n:Met
The counts add up to the marks
A second look by the AI
Meets:The numbers on the Result line match the marks, and the table has one row for each criterion not marked met.
Does not meet:A count on the Result line is wrong, or a criterion not marked met has no row, or a met criterion has one.
Claude:Met
ChatGPT:Met
n8n:Met
Each mark and each fix follows the rubric's own words
A second look by the AI
Meets:Each mark matches the rubric's wording for that level, and each fix uses only the rubric's wording and facts already in the work.
Does not meet:A mark gives credit the rubric does not allow, or a fix suggests content that is in neither the rubric nor the work.
Claude:Met
ChatGPT:Not met
n8n:Met
You decide whether you agree with each mark
You check this one
Meets:You read the evidence beside each mark, decided whether you agree, and treated the marks as a first read.
Does not meet:The marks are used as a final grade without being read against the work.
Claude:For your read
ChatGPT:For your read
n8n:For your read
Claude: met 8 of 8 (including 1 for your read)
ChatGPT: met 7 of 8 (including 1 for your read)
n8n: met 8 of 8 (including 1 for your read)
Learn from this template
A scoring guide takes one criterion at a time, so each mark rests on a passage quoted from the work and on the rubric's own words. Marks come from a short closed list, and the last check is yours: you decide whether you agree with each one.
Where to see it: step 1 (List each criterion and what met looks like); step 2 (Find the passage for each criterion); step 3 (Mark each criterion).