Matlab Plotting Skill

MATLAB Plotting Skill

Overview

Use this skill to turn user data and a plotting goal into a rendered MATLAB figure. The bundled MATLAB assets include 51 plotting schemes, a data-schema inference layer, palette defaults, exporters, and a report writer.

This skill is self-contained. Do not rely on private local folders, external plotting repositories, encrypted MATLAB files, or article image packs. If local plotting folders are present, use them only to identify chart tasks and read docs/local-resource-intake.md before turning any idea into public code.

Workflow

1. Inspect the user's request and identify the communication task: trend,

relationship, matrix, comparison, distribution, ranking, composition, multivariate, spatial/vector, or layout.

1. If the user provided a file, check whether it is CSV, Excel, or MAT. See

references/data-contract.md when the data shape is unclear.

1. Choose the plot by reading references/scheme-catalog.md only when the

built-in name/tag mapping is not enough. To inspect one scheme without opening MATLAB, use --scheme-info , --scheme-info --status, or --scheme-info-json --status. Read references/example-prompts.md when writing usage examples for a user. If the idea came from a private local plotting folder, write it as a public communication task first and check docs/local-resource-intake.md before touching code.

1. When the user needs to inspect file structure before choosing a plot, use

--inspect-data, which calls mpInspectData and returns schema JSON.

1. When the user wants to inspect the choice before rendering, preview the

selection with --plan-only, which calls mpPlan and returns JSON:

``bash skills/matlab-plotting-skill/scripts/render_with_matlab.sh --plan-only --data --goal "" ``

Use --explain instead of --plan-only when the user wants a concise terminal explanation rather than JSON.

1. Prefer the bundled CLI for rendering:

``bash skills/matlab-plotting-skill/scripts/render_with_matlab.sh --data --goal "" --out --formats png,svg ``

Use --scheme when the user asks for a specific scheme.

1. If MATLAB is missing, explain that the user needs MATLAB CLI and provide the

exact command to run later. Do not claim a figure was rendered.

1. After rendering, report the selected scheme, output files, and any quality

warnings from render_report.md. Use render_report.json when a script or follow-up automation needs structured metadata. Read docs/cli-output-contract.md before depending on JSON fields.

For a first-time user, mirror the sequence in docs/first-render-walkthrough.md: run scripts/doctor.sh or render_with_matlab.sh --doctor --out when the checkout or installed Skill is new or confusing, check MATLAB, inspect data, run --plan-only, render PNG/SVG, then review the report and privacy checks. Use docs/troubleshooting.md when MATLAB cannot be found, a render hangs, no figure files appear, or the selected scheme looks wrong.

When writing user-facing summaries, reports, or documentation snippets, keep the tone grounded, maintainer-style, and specific. Follow docs/writing-style.md: say what works today, name the MATLAB boundary, and avoid unsupported adoption or approval claims.

MATLAB CLI

The renderer uses MATLAB_BIN when set, then falls back to matlab on PATH. MATLAB-backed commands have a 600-second timeout by default. --smoke-test auto-scales the timeout from the catalog size and prints the applied budget; set MP_MATLAB_TIMEOUT_SECONDS=0 only for deliberate long local renders. On macOS, a common value is:

MATLAB_BIN=/Applications/MATLAB_R2025a.app/bin/matlab

Read references/matlab-cli.md for troubleshooting. Read docs/private-data-handling.md before asking a user to share logs or reports from private data.

Selection Defaults

• Time column plus numeric values: trend schemes.
• Two numeric columns: scatter or density scatter.
• Numeric matrix: heatmap schemes.
• Category plus numeric values: grouped bar, box/jitter, or ranking schemes.
• Percent or part-whole language: composition schemes.
• Positive and negative values: diverging bar or positive-negative area.
• "Zoom", "event", "detail", or "local": zoomed inset line.
• "Paper", "panel", "overview", or "layout": multi-panel or annotated layout.
• "Spectrum", "impedance", "power", "energy", "harmonic", "block", or

"workflow": check whether the public catalog already has a matching scheme; if not, describe the closest existing option and record a backlog task rather than importing a private local template.

When a MAT file has multiple plausible variables, list the candidates and ask the user which variable to render instead of guessing.

Safety Rules

• Keep generated scripts and outputs in the user's chosen project directory.
• Do not write into the skill directory during normal use.
• Do not copy raw private data into this repository.
• Treat local plotting collections as requirements sources only; do not copy

their code, labels, screenshots, watermarks, binary artifacts, or author metadata.

• Do not introduce .p, .fig, committed .mat, .docx, .pdf, article

screenshots, local absolute paths, email addresses, or provenance-unclear source files.

• Keep generated preview assets out of skills/, scripts/, and source or

template directories; use reviewed output or gallery folders.

• Use synthetic data for smoke tests and examples.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Kkkakania
• Source: Kkkakania/matlab-plotting-skill
• License: MIT

Install and usage instructions live in the source repository linked above.