Skip to content

matlab/agent-skills-playground / demos/assessment-generation-for-matlab-grader/skills/matlab-generate-grader-assessments

matlab-generate-grader-assessments

Verified

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.

references/tests-prompt.md

View on GitHub
# MATLAB Grader assessment configuration prompt

Create `assessments.md`, not a padded test script. Start with a **Student Template Line Locks** section, then add a requirement-to-assessment matrix, then include exact MATLAB Grader setup directions for every row. Add an `Optional feedback on incorrect submission` column. Use `—` when no feedback is warranted; otherwise provide feedback tied to a validated incorrect variant.

Resolve the content language from the course profile. Write learner-visible assessment names, expected-evidence prose, MATLAB Code comments, and optional feedback in the resolved content language. Keep MATLAB Grader test type labels, instructor-facing setup headings, UI field names, MATLAB code, identifiers, function signatures, class names, variable names, file names, and MATLAB keywords unchanged.

In **Student Template Line Locks**, list the final 1-based `template.m` line numbers that instructors should lock in MATLAB Grader after pasting the student template. Derive the line numbers from the final generated `template.m`. Use a table with columns `Line(s)`, `Lock?`, `Exact template text`, and `Reason`. Use inclusive ranges for contiguous locked lines and comma-separated groups only when needed. For a single line, quote the exact line text; for a range, quote the first and last line text. Lock required scaffold comments, fixed signatures, fixed class or inheritance declarations, required property or method signatures, provided setup values, structural `end` lines, and referenced-file usage scaffolding that must remain intact. Do not lock learner implementation placeholders or blank lines intended for student work. If no lines should be locked, write `No template lines need to be locked for this item.`

Use **Variable equals reference solution** only for ordinary equality of one student variable in a Script submission. It cannot compare a vector, expression, or list of multiple variables, and it is not available for Function or class-submission items. Use **MATLAB Code** for a custom property that direct equality cannot check, whenever one assessment must compare multiple student variables, for every Function-submission output check, and for class-submission checks.

For **Variable equals reference solution** rows, the `assessments.md` matrix may describe the MATLAB Grader UI field as `Variable name: <name>`, but the row's `Code to paste` value must be `—` because there is no MATLAB assessment code. In combined `AllGraderItems.md` output, the fenced `Copy` block for that assessment field must contain only `<name>` as a bare MATLAB variable identifier, not `Variable name: <name>`.

Each Function assessment must assign its inputs, call both the learner function and its `reference.<functionName>` counterpart, then call `assessVariableEqual` on the learner output and reference output. Do not assume `function_call.m` variables or `referenceVariables` are available, and do not reproduce the solution algorithm. When class matters, compare to the class of the reference-function output.

For Class Definition, Class Inheritance, and Class Methods items, write MATLAB Code assessments that instantiate the learner class when concrete, instantiate `reference.<ClassName>` when needed, and compare observable class behavior. Use `superclasses(obj)` for inheritance, `properties(obj)` for required properties, `methods(obj)` for methods, and object properties for constructor defaults or method-updated state. For method tests, run the same referenced-file setup and method calls on learner and reference objects, then compare resulting properties with `assessVariableEqual`.

For Object Usage Script items, assess variables created by the learner script. Use `exist`, `class`, and object property comparisons against `referenceVariables.<name>` for clear feedback. Use whole-variable **Variable equals reference solution** only when one object variable equality is truly the intended evidence and targeted property checks are not needed.

For Script submissions, **Variable equals reference solution** allows ±0.1% relative tolerance or ±0.0001 absolute tolerance between the learner and reference values. Use **MATLAB Code** with `assessVariableEqual` and the full parameter names `RelativeTolerance` or `AbsoluteTolerance` to override that default. Never generate `RelTol` or `AbsTol`; those shortened names are invalid for this MATLAB Grader assessment code.

Use **Function or Keyword is present** only for an explicitly required named construct. Use **Function or Keyword is absent** only for an explicitly prohibited shortcut. Do not add construct tests merely to make a suite longer.

For every assessment, include a concise learner-visible assessment name in the MATLAB Grader UI fields. For Function or Keyword presence/absence assessments, the name must describe the behavior being assessed and must not include the command or keyword checked by that row. Put the checked command or keyword only in its configuration field.

Each row must identify its learning-objective evidence, UI field values, code to paste when applicable, expected evidence, optional feedback, and traceability. Feedback for formative items may identify a next check; feedback for summative items must diagnose only and must not reveal an answer or implementation approach. Reject duplicate checks. Generate `tests.m` only when at least one row is MATLAB Code; each section then corresponds to exactly one MATLAB Code row.

List every referenced `.m` or data file needed by assessments. Generate helper checks as readable `.m` files only; do not generate `.p` files. If an educator wants hidden helper logic, state that they may manually pcode a reviewed helper `.m` before uploading it to MATLAB Grader.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 323kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 323kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 323kUpdated Claude CodeCodexVerifiedInstall