/matlab-generate-grader-assessments

name: matlab-generate-grader-assessments
description: Generate MATLAB Grader assessment items that are suitability-gated, profile-driven, feedback-aware, and validated through MATLAB MCP. Produces Script, Function, Class Definition, Class Inheritance, Object Usage, and Class Methods items with MATLAB Grader assessment setup instructions.
license: https://www.mathworks.com/content/dam/mathworks/license/pmrl/license.md
metadata:
  author: MathWorks
  version: "2.0"

MATLAB Grader Assessment Item Generator

Generate complete, reviewable MATLAB Grader assessment items from an observable MATLAB-code learning objective. This skill supports **Script**, **Function**, **Class Definition**, **Class Inheritance**, **Object Usage**, and **Class Methods** items.

An item folder contains `description.txt`, `solution.m`, `template.m`, and `assessments.md`; for Function, Class Definition, Class Inheritance, and Class Methods items, it also contains `function_call.m`. `tests.m` is created only when the item needs one or more **MATLAB Code** assessments. Referenced class, helper, and data files must be human-readable source or data files and must be documented in `assessments.md`; do not generate `.p` files.

Staged loading

Read `matlab-grader-course-profile.md` in the course-material root before proposing an item. If it is absent, complete the one-time setup below first.

Load reference files only for the current stage:

| Stage | Load | | --- | --- | | Suitability and proposal | `references/assessment-item-types.md`, `references/options-prompt.md` | | Student artifacts | `references/description-prompt.md`, `solution-prompt.md`, `template-prompt.md` | | Function and class run artifacts | `references/function-call-prompt.md` | | Assessment configuration | `references/tests-prompt.md` | | Assessment rationale or feedback design | `references/assessment-research.md` | | QTI export | `references/qti3-prompt.md` | | Combined output | `references/all-grader-items-template.md` |

Do not preload references for later stages.

MATLAB MCP is mandatory. Confirm that a working MATLAB MCP session can run a small MATLAB command before proposing or generating an item. If it cannot, stop and say that generation is blocked; do not claim that code or assessments have been validated.

Use `matlab-read-doc` when current MATLAB Grader behavior or MATLAB syntax needs verification. Use `matlab-debugging` only to diagnose a failed MATLAB MCP validation.

One-time course profile setup

Create `matlab-grader-course-profile.md` with YAML front matter and readable Markdown body. Ask explicit questions for only missing values; never present “press Enter for defaults.”

The front matter must include:

---
profile_version: 1
output_location: assessments
assessment_purpose: summative
qti3_export: false
require_matlab_mcp: true
content_language: auto
language_policy:
  comments: match_content_language
  feedback: match_content_language
  assessment_names: match_content_language
  instructor_setup: english
  preserve_code_identifiers: true
coding_practice_progression:
  enabled: true
  low: [descriptive names, string literals for new text]
  moderate: [descriptive names, string literals for new text, avoid unsafe dynamic-workspace functions]
  high: [descriptive names, string literals for new text, avoid unsafe dynamic-workspace functions, concise functions]
learning_objectives:
  - objective: "..."
    allowed_complexity: [low, moderate]
    preferred_submission: Script
---

Document that `low`, `moderate`, and `high` are the only complexity labels. The coding-practice progression is an authoring gate and learner guidance; it is not a student scoring criterion unless a future objective explicitly makes it assessable.

`content_language` controls student-facing language. Use `auto` to infer the resolved content language from the approved problem description or task statement for each item. Use a BCP 47 or ISO-style language code such as `en`, `es`, or `ko` to force all generated items to that language. If an older profile omits `content_language`, default to `auto`. When auto-detection is ambiguous or the prompt mixes languages, use the dominant language of the student-facing problem description.

Apply `language_policy` after resolving the content language. Generate student-facing prose, code comments, learner-visible assessment names, and optional feedback in the resolved content language. Preserve MATLAB code, identifiers, function signatures, class names, variable names, file names, MATLAB keywords, MATLAB Grader test type labels, and instructor-facing setup headings unless the user explicitly asks to localize them.

Suitability and proposal gate

Before generation, determine whether the objective has observable MATLAB-code evidence: a workspace variable, function output, required construct, prohibited shortcut, or input-contract behavior.

• If it does not, stop. Explain why MATLAB Grader cannot directly observe the objective, propose an assessable rewording, or recommend another modality such as a written explanation, design review, or manual rubric.
• If it does, recommend Script, Function, Class Definition, Class Inheritance, Object Usage, or Class Methods and give a brief rationale.
• For learner-authored `classdef` work, warn that class definitions must be submitted as plain `.m` files, not Live Script `.m` or `.mlx` files. The submitted class must be concrete whenever assessments instantiate it.
• Do not generate a directly auto-graded learner-authored abstract-class item when validation depends on instantiation. Abstract classes are allowed as referenced superclasses; recommend a concrete subclass, object-usage task, manual review, or design rubric when the learner's submitted class itself is abstract.
• Read the objective’s allowed complexity from the profile. If the requested level is not supported, report that it is unsupported and do not add unrelated requirements to inflate complexity.

For