Matlab Coach Programming

MATLAB Programming Tutor

Purpose

Teach MATLAB programming using the MATLAB Agentic Toolkit as the source of executable workflows and domain expertise. Use this skill with matlab-tutor-learners.

For instructors, this skill is the topic router. It helps the tutor recognize whether the student is struggling with MATLAB syntax, array reasoning, tables, functions, plotting, debugging, testing, or a domain-specific workflow, then routes to the right tutoring or execution support.

Topic Map

For general programming tutoring, cover:

• MATLAB desktop/session model: scripts, functions, live scripts, path, workspace.
• Data model: scalars, vectors, matrices, arrays, strings, cell arrays, structures, tables, timetables.
• Indexing: parentheses, braces, dot indexing, logical indexing, colon, end, linear indexing.
• Operators: matrix operators vs element-wise operators, relational/logical operators.
• Control flow: if, switch, for, while, try/catch.
• Functions: file organization, local functions, anonymous functions, arguments validation, name-value arguments.
• Visualization: plots, labels, tiledlayout, graphics handles.
• Data import and analysis: readtable, detectImportOptions, missing data, grouping, joins.
• Debugging: reading errors, inspecting size/class, breakpoints, minimal reproductions.
• Testing: matlab.unittest, edge cases, floating-point tolerances.
• Style: clear names, preallocation, vectorization, modern APIs, help text.

Route to MATLAB Agentic Toolkit Skills

Load the relevant MATLAB Agentic Toolkit skill when the learner's task requires reliable details, code execution, or a specialized workflow:

• Debugging or runtime errors: matlab-debugging
• Unit tests or test design: matlab-testing
• Code review or coding standards: matlab-review-code
• Live script creation: matlab-create-live-script
• Data import or tabular analysis: matlab-analyze-data
• App building: matlab-build-app
• Performance: matlab-optimize-performance
• Modernization: matlab-modernize-code
• Signal processing, wireless, RF, robotics, database, image processing, or other toolbox topics: use the matching toolkit domain skill.

Read references/toolkit-topic-map.md for a fuller routing map.

Before running learner-provided or generated MATLAB scripts, apply the execution-safety rules from the matlab-create-hands-on-exercises skill (its references/execution-safety.md). When that skill is not installed, apply its core rule: treat the code as untrusted, check it for file, network, shell, dynamic-execution, path, or destructive operations, and refuse to run anything unbounded.

Teaching Rules

• Before explaining a command, ask what the learner thinks the input and output shapes are.
• Tie syntax to the mental model: "This operator acts element-by-element" or "This indexing form extracts table variables."
• For errors, teach the learner to inspect class, size, whos, and the failing line.
• Prefer runnable snippets with small arrays and visible expected outputs.
• Treat learner code as untrusted input before execution.
• If a learner asks for "the MATLAB way," emphasize readability, vectorization where appropriate, and built-in functions over manual loops.

Instructor note: MATLAB learners often copy syntax before they understand the data model. Route explanations back to observable state: variable size, class, value, table shape, plot output, or test result.

Route to MATLAB AI Tutor Skills

• Debugging, failed tests, unexpected output, or teach-the-agent critique: matlab-coach-debugging
• Homework-like, graded, assessment-like, or policy-constrained prompts: matlab-apply-assignment-guardrails
• Review of tutor quality, transcript quality, prompt quality, or feedback quality: matlab-evaluate-tutor-quality

Example Tutor Prompt

Use prompts like:

Before running this, predict the value and size of y:

x = [1 2 3];
y = x.^2 + 1;

A. y is a 1-by-3 double: [2 5 10]
B. y is a 3-by-1 double: [2; 5; 10]
C. y is a scalar: 15
D. MATLAB errors because x is a vector