Mark student writing against a rubric
Marks a batch of student writing against the teacher's rubric with criterion scores, quoted evidence and one next step each, and flags borderline scripts for teacher review.
Rubric marking is reliable when every score is tied to the wording of a descriptor and to evidence in the script, when the same reading of each descriptor is applied to every script, and when uncertain cases go to a human. It goes wrong through drift (scripts marked later are judged against earlier ones instead of the rubric), halo effects (strong spelling raises the score for ideas), length bias, and feedback that names a level without telling the student what to do next. The teacher stays the assessor: this pass gives a consistent first marking with the evidence laid out, so the teacher can moderate quickly and spend their time on the scripts that genuinely need judgement.
Mark the scripts below against the rubric for this taskOnly if [GRADE_LEVEL] is given: ().
- Read the rubric and write Marking notes: for each criterion, what evidence distinguishes adjacent levels, in one or two lines. If a descriptor is ambiguous, state the interpretation you will apply to every script. If the rubric is missing level descriptors or the scripts are not separated, stop and ask for what is missing.
- Read all the scripts once before scoring any of them.
- Score criterion by criterion across all scripts, not script by script, to keep each criterion consistent. For every score, quote the shortest piece of the script that justifies it (or note what is absent).
- Judge each criterion on its own. Do not let spelling and grammar affect content criteria unless the rubric says so, and do not reward length for its own sake.
- For each script, write one next step: specific, actionable in the next piece of work, and phrased to the student ("In your next paragraph, add a quotation that shows… and explain how it proves…").
- Flag a script for teacher review when: a criterion sits on the boundary between two levels; the script does not answer the task; it is far shorter than expected or unfinished; its style or level is far from the rest of the set or from what the task and grade would lead you to expect (say only that it differs and that the teacher may want to compare it with the student's other work; do not accuse anyone of copying or AI use); or it contains anything that suggests a student may be at risk of harm (flag it as a welfare concern for the teacher to follow up through the school's safeguarding procedure the same day, and do not mark that script).
- Summarise patterns across the class: the criteria most and least secure, and one suggestion for whole-class reteaching.
- Scores are provisional and for the teacher to confirm; say so once in Marking notes.
- Use only the rubric's levels and scale. Never invent extra criteria or a grade conversion the teacher did not give.
- Every score has a quotation or an explicit "not present". Quotations are copied exactly.
- Refer to students by the script code only. If a full name appears, use the code instead.
- Do not judge students' identities, dialects or home languages; apply the rubric to the writing.
Marking notes
Interpretations of each criterion and the provisional-score reminder.
Score summary
Table: Script | one column per criterion | Total (if the rubric totals) | Flag.
Script feedback
For each script: a table Criterion | Score | Evidence (quote) | Why this level, then "Next step:" one sentence to the student.
Flagged for teacher review
Script | Reason | What to check.
Class patterns
Strongest and weakest criteria with counts, and a reteaching suggestion.
3 required values still a placeholder; the assistant will ask for them.
details
- kind
- Prompt: a task you run by name to get one finished thing back
- domain
- Learning and education
- category
- Teaching
- level
- Intermediate
- made for
- Teacher / tutor
- risk
- read-only
- version
- v1.0.0 · incubating
- reviewed
- 2026-10-03
- works in
- Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Antigravity, OpenCode, Windsurf, Zed, Continue, AGENTS.md, ChatGPT, claude.ai
use in
npx @hermes-hq/hodios install mark-student-work-against-rubric --target claude-codeThis entry is in the full catalog, not the curated set the skills installer and plugins carry, so install it with the Hodios CLI.
pairs well with
All of TeachingCreate an analytic rubric
Builds an analytic rubric with distinct criteria, performance levels and observable descriptors aligned to learning objectives, plus scoring notes. Use when setting or marking an assignment.
create-rubricBuild a feedback comment bank
Builds a reusable bank of specific feedback comments keyed to rubric criteria and levels, each with a next step, plus whole-class feedback, to speed up marking a class set.
build-feedback-comment-bankAnalyse a class's assessment results
Analyses a class's scores by item and standard to find the weakest skills, likely misconceptions, suspect items and reteaching groups. Use after marking a test or quiz.
analyze-class-assessment-resultsDesign self- and peer-assessment
Designs self-assessment and peer feedback protocols tied to success criteria, with sentence starters, a feedback form and steps that make students act on the feedback.
design-self-and-peer-assessmentAdapt a lesson for a deaf pupil
Adapts a lesson for a deaf or hearing-impaired pupil with seating, lighting, radio aid use, captions, visual vocabulary, pre-teaching, check-ins and rules for group talk.
adapt-lesson-for-hearing-impaired-pupilAdapt a text to several reading levels
Rewrites a passage at several reading levels while keeping the key content and target vocabulary, with a glossary and comprehension questions for each version. For teachers with mixed-ability classes.
adapt-text-reading-level