agentleFS
AxectAxect/skills/scienceplot-py/SKILL.md

scienceplot-py

Skill4 starsChanged 20 days agoMITView on GitHub

Generate Python matplotlib plotting scripts in the user's scienceplots lab style: with plt.style.context(["science", "nature"]), pparam = dict(...), raw-string LaTeX labels, and fig.savefig(..., dpi=300, bbox_inches='tight'). Support parquet, CSV, NPY/NPZ data and single-line, multi-line, scatter/errorbar, or subplot variants. Generate both the reproducible plotting script and the requested PNG/PDF/SVG figure by default. Use for publication-style plot scripts, science/nature figures, plot-from-data scaffolds, and Korean or English requests for matplotlib graph scripts.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 2 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Every generated script MUST keep these load-bearing patterns intact.

    line 29
  2. Very specific:

    Default filename is plot.png.

    line 47
  3. Very specific:

    Write the .py file with the Write tool.

    line 71
  4. Very specific:

    Prefer uv run <path> when the current project is the active uv project.

    line 75
  5. Very specific:

    Verify that every requested output file exists and is non-empty.

    line 81
  6. Distinctive:

    If the user explicitly requests script-only, stop after step 5 and return an execution hint instead of running the script.

    line 89

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

scienceplot-py — Matplotlib Plot + Figure Generator (scienceplots / science+nature)

Generate a Python plotting script that always follows the user's lab template shape, execute it, and verify the rendered figure artifact. The skill returns both the .py path and the generated figure path(s). The default workflow is script plus figure generation; use script-only mode only when the user explicitly requests a script without execution.

The canonical template lives at ~/Socialst/Templates/PyPlot_Template/pq_plot.py. Every variant in this skill is a structural extension of that file.

Mandatory style invariants

Every generated script MUST keep these load-bearing patterns intact. Do not "clean up" any of them, even if they look redundant or unused.

  1. import scienceplots — required. It registers the science and nature styles by import side-effect; the symbol is never referenced directly. Linters will flag it as unused — keep it anyway.
  2. Style context block — all plotting code lives inside with plt.style.context(["science", "nature"]):. Never substitute plt.style.use(...) or call any plotting outside this block.
  3. pparam dict — axis configuration is a dict, applied with ax.set(**pparam). Do not inline ax.set_xlabel(...), set_ylabel(...), set_title(...) calls when pparam would do.
  4. ax.autoscale(tight=True) — called on every axis, before ax.set(**pparam). For subplots, loop over all axes.
  5. Raw-string LaTeX — every label, title, legend entry uses r'...'. Non-raw '$x$' works for the literal $x$ but breaks the moment a backslash appears (\alpha, \sigma, \mathrm{...}). Always r-prefix.
  6. savefig — fig.savefig(<path>, dpi=300, bbox_inches='tight'). Default filename is plot.png. PDF/SVG output is fine if asked, but keep dpi=300 and bbox_inches='tight'. (Bump higher only if the user explicitly asks for it — the lab default is 300.)

Workflow

  1. Gather what the user has not already given:
    • Data source: parquet / CSV / .npy / .npz, plus the file path
    • x / y column or array names (and y-error if errorbar)
    • Plot variant: single line, multi-line + legend, scatter / errorbar, or subplots (rows × cols)
    • Axis labels, title, legend labels (LaTeX OK — strings will be wrapped in raw-string form)
    • Output path (default: plot.png next to the data file or in cwd)
  2. Pick the matching template under references/:
    • single_line.py — one series, one ax (mirrors pq_plot.py)
    • multi_line.py — multiple series, one ax, with legend
    • scatter_errorbar.py — ax.errorbar(...) (or ax.scatter(...))
    • subplots.py — multi-panel fig, axes = plt.subplots(rows, cols)
  3. Swap in the requested data-loader block. See references/data_loaders.md for parquet / CSV / .npy / .npz snippets.
  4. Substitute column names, label strings, title, output filename. Preserve every invariant from the section above.
  5. Write the .py file with the Write tool. Preserve the requested output location in the script rather than relying on an accidental current working directory.
  6. Render the figure by executing the generated script from the project root. Prefer uv run <path> when the current project is the active uv project; when the project environment is in a subdirectory, use uv run --project <project-dir> python <script-path>. Do not install packages automatically: if scienceplots, matplotlib, pandas, or numpy is missing, report the dependency error and the command needed after the user fixes the environment.
  7. Verify that every requested output file exists and is non-empty. For report figures, generate the requested raster output (normally PNG) and a PDF or SVG companion, each with dpi=300 and bbox_inches='tight'. If the figure is available to the harness, inspect it for clipped labels, missing glyphs, empty axes, and accidental transparent backgrounds.
  8. Return the script path, generated figure path(s), the exact render command, and a short render-status summary.

If the user explicitly requests script-only, stop after step 5 and return an execution hint instead of running the script.

Data sources

SourceImportsRead call
Parquetimport pandas as pddf = pd.read_parquet('data.parquet')
CSVimport pandas as pddf = pd.read_csv('data.csv')
NumPy .npyimport numpy as npdata = np.load('data.npy') (then column-slice)
NumPy .npzimport numpy as npdata = np.load('data.npz'); x = data['x']

Full snippets in references/data_loaders.md.

Templates

Each reference is a self-contained, runnable skeleton matching the mandatory style invariants. Read the file, adapt strings/columns in memory, then Write the result to the user's chosen path.

VariantFile
Single line (base)references/single_line.py
Multi-line + legendreferences/multi_line.py
Scatter / errorbarreferences/scatter_errorbar.py
Subplots (multi-panel)references/subplots.py
Data loadersreferences/data_loaders.md

What this skill does NOT do

  • It does not install scienceplots / matplotlib / pandas / numpy. Assume they are already available in the target environment; dependency installation remains the user's or project's responsibility.
  • It does not produce methodology / architecture / pipeline diagrams. For those, use wide-slide-illustrator or handdrawn-schematic.
  • It does not change the style invariants. If the user asks for a different style (e.g., seaborn, ggplot, plain matplotlib), tell them this skill is specifically for the science+nature template and offer to write the alternative as a plain script outside the skill.
  • It does not silently treat a successful Python exit as proof of a valid figure: output existence and basic visual sanity checks are part of the rendering workflow.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.