Py Math Viz

Create clear math/data visualizations in Python and export to PNG (matplotlib/seaborn; optional OpenCV post-processing). Use when the user asks to plot funct...

Security Scan

The skill's code, instructions, and requirements are consistent with a Python plotting utility; included scripts and a weather fetch helper match the stated purpose and don't request unrelated credentials or installs.

Purpose & Capability

Name/description (plot PNGs with matplotlib/seaborn and optional OpenCV) align with the bundled scripts: plot_from_spec.py (JSON→PNG renderer), tile_images.py (OpenCV image tiling), and weather.py (fetches weather + renders a 24h sprite). The weather script is an expected convenience example for plotting.

Instruction Scope

SKILL.md instructs running the provided scripts and writing outputs to out/. plot_from_spec.py reads CSVs or inline arrays from a spec you provide (so it will read any file paths you supply). weather.py makes an outbound HTTP call to api.open-meteo.com to fetch forecast data. These behaviors are coherent with plotting, but the scripts will read and write files you point them at and the weather helper performs a network request — review inputs you pass and expect network traffic when using the weather script.

Install Mechanism

No install spec or remote download; this is an instruction/code-only skill. The SKILL.md notes a specific virtualenv path and expected Python packages (matplotlib, seaborn, numpy, opencv-python-headless), which is a runtime assumption rather than an installer fetching arbitrary code.

Credentials

The skill requests no environment variables, no credentials, and no config paths. There are no hidden secrets requested; external network access is limited to the open-meteo API called by weather.py (expected for a weather plotting helper).

Persistence & Privilege

always is false and the skill does not modify other skills or system-wide settings. It only writes output files to paths you or the instructions specify (out/ by default).

Assessment

This skill appears to be what it says: a small Python plotting toolkit with an OpenCV tiler and a weather-example that fetches data from api.open-meteo.com. Before using: (1) confirm you trust the source — the full scripts are included so you can inspect them (good practice); (2) run inside an isolated environment/venv or sandbox (SKILL.md references /root/.openclaw/workspace/.venv-math/bin/python but you can use any Python with the listed libs); (3) only pass CSV/image file paths you intend to expose — the scripts will read any path you provide; (4) expect the weather script to perform outbound HTTP requests (no credentials required); (5) there are no required secrets or installers embedded, so risk is typical for running third-party scripts (review and sandbox if you have sensitive data).

Use Python to produce clean, readable plots and export to PNG by default.

Assumptions (this workspace):

• Preferred interpreter: /root/.openclaw/workspace/.venv-math/bin/python
• Libraries: matplotlib, seaborn, numpy (already), opencv-python-headless (installed)

Defaults (quality rules)

• Output: PNG, dpi>=200, white background
• Always label axes; add title if it helps
• Use legend when multiple series exist (and place it so it doesn’t cover data)
• Avoid overlap:
  • prefer constrained_layout=True and/or plt.tight_layout()
  • rotate dense x-ticks, shorten tick labels
  • increase figure size instead of cramming
• Use consistent styling:
  • seaborn.set_theme(style="whitegrid")
  • colorblind-friendly palette if unsure

Quick start (script)

Render a plot from a JSON spec. To keep the workspace root clean, write outputs into out/ (or out/_scratch/ for throwaway experiments):

/root/.openclaw/workspace/.venv-math/bin/python \
  skills/py-math-viz/scripts/plot_from_spec.py \
  --spec spec.json --out out/plots/plot.png

Then send out/plots/plot.png (Telegram: use the message tool with filePath).

Plot spec

The spec is intentionally small. Read:

• references/spec.md — schema + examples
• references/quick-recipes.md — recipes index (points to JSON specs / seaborn snippets / image tiling)

OpenCV helpers

Tile multiple PNGs into one (for comparisons):

/root/.openclaw/workspace/.venv-math/bin/python \
  skills/py-math-viz/scripts/tile_images.py \
  --out tiled.png img1.png img2.png img3.png

When NOT to use

• Interactive dashboards (Plotly/Dash) unless explicitly asked
• 3D/mesh-heavy rendering (use specialized tools if needed)