Engineering Figure Agent

Use this skill after the figure goal is already reasonably clear.

Boundary

Good fit:

• Turn a figure brief into a conceptual diagram, workflow figure, schematic, or exact publication plot.
• Choose between image, plot, and mixed mode.
• Build prompts, render plots, and refine existing briefs.

Not the main tool for:

• Deciding from scratch what claim the paper should visualize.
• Full paper-argument or reviewer-style critique.

If the user is still deciding the figure claim, panel logic, or caption argument, use an available research-writing or paper-analysis skill upstream first.

Core Decision

• image mode: conceptual figures, architecture diagrams, algorithm workflows, graphical abstracts, schematics, reference-inspired redraws.
• plot mode: exact bar charts, trend curves, heatmaps, scatter plots, ablation plots, benchmark summaries.
• mixed mode: render quantitative panels locally first, then use image generation only for conceptual panels.

Never use image generation for exact values, axes, or benchmark geometry.

Default Workflow

1. Inspect the user input and decide whether a figure brief is already present.
2. If needed, create a brief using references/figure-brief-spec.md.
3. Choose image, plot, or mixed mode.
4. For conceptual figures, prefer the prompt-builder scripts; read template references only when you need to inspect or customize template wording.
5. For exact plots, create a concise plot request and render it locally.
6. Keep labels short, claims source-grounded, and outputs publication-oriented.
7. Save prompt/spec/output paths when files are produced.
8. Before finishing, run only the checks needed for the touched path.

Reference Loading

Read only what is needed:

• references/figure-brief-spec.md: brief structure and mode rules.
• references/publication-figure-design.md: publication styling defaults.
• references/provider-selection.md: Gemini/Banana/OpenAI setup.
• references/highres-policy.md: high-resolution and fail-closed rules.
• references/chinese-labels.md: Chinese label readability rules.
• references/engineering-figure-templates.md and references/materials-science-figure-template.md: template wording when script output needs customization.
• references/publication-chart-patterns.md, references/natural-language-plot-workflow.md, and references/publication-plot-api.md: exact or plot-heavy workflows.
• references/editable-figure-handoff.md: optional editable SVG handoff.

Scripts

• Build a prompt without network calls:

python scripts/build_engineering_figure_prompt.py --figure-template system-architecture --lang en "technical background"
python scripts/build_materials_figure_prompt.py --materials-figure mechanism-figure --lang en "scientific background"

• Generate or edit a conceptual image:

python scripts/generate_image.py --provider openai --figure-template system-architecture --lang en "technical background"
python scripts/generate_image.py --provider openai --materials-figure mechanism-figure --lang en "scientific background"

• Render an exact plot:

python scripts/build_plot_spec.py request.json --out spec.json
python scripts/plot_publication_figure.py spec.json --out-path output/figure --formats png pdf svg

• Use the unified CLI:

python scripts/efa.py prompt --figure-template system-architecture "technical background"
python scripts/efa.py prompt --materials-figure mechanism-figure "scientific background"
python scripts/efa.py plot request.json --out-path output/figure
python scripts/efa.py check

Quality Rules

• Do not fabricate measurements, benchmark values, hardware specs, or unsupported causal claims.
• Keep arrows, reading order, and module hierarchy explicit.
• Favor white backgrounds, publication-style spacing, and readable labels.
• Use prompt-first workflows when layout fidelity matters.
• For high-resolution requests, follow references/highres-policy.md.
• For third-party providers, require explicit opt-in before sending keys or user files.
• For Chinese figures, preserve standard technical notation when it improves clarity.