Skip to content

gpsaggese/helpers / .claude/skills/figure.create_tikz

figure.create_tikz

Verified

Generate a TikZ code for an image or a description

FileKindSize
SKILL.mdskill md3.5 KB
---
description: Generate a TikZ code for an image or a description
model: sonnet
---

# Goal
- Convert images or text descriptions into publication-quality TikZ LaTeX code
- Generate a compilable LaTeX file, render it to PNG, and refine the output until
  it matches the input

## When to Use
- Use this skill to:
  - Create publication-quality diagrams, plots, or visualizations
  - Convert hand-drawn sketches or images into reproducible TikZ code
  - Generate diagrams for LaTeX documents

## When NOT to Use
- Do not use this skill for:
  - Complex photographs needing photorealistic rendering
  - Plots from large datasets (use dedicated plotting libraries)
  - Advanced 3D visualizations

# Workflow

## Receive and Understand the Input

- You receive either:
  1. An image of a diagram
  2. A text description of a concept or situation

## Choose Figure Type

- Pick the diagram structure that fits the input:
  - **FLOWCHART**: sequential steps, decision branches, pipelines
  - **STRUCTURAL**: containment (things inside other things), architecture
  - **ILLUSTRATIVE**: plots, physical cross-sections, or abstract spatial metaphors

## Generate TikZ Code
- Generate valid LaTeX with the TikZ package, as a complete minimal working
  example:
  ```latex
  \documentclass{standalone}
  \usepackage{tikz}
  \begin{document}
  \begin{tikzpicture}
  ...
  \end{tikzpicture}
  \end{document}
  ```

**Preserve layout accurately**

- If converting from an image `<IMAGE>`, reproduce the layout precisely
- Preserve proportions, relative positions, and symmetry
- Use coordinates and scaling where appropriate
- Approximate complex curves with TikZ paths when needed

**Follow the TikZ conventions**

- Follow `.claude/skills/tikz.rules.md`, especially Colors, Typography, Layout,
  Reusable Patterns, and Polish

**Handle ambiguity**

- Make reasonable assumptions about unclear inputs
- Prioritize clarity and visual correctness over perfection

## Save the File
- Save the LaTeX code to `./tikz_figure.tex` in the current directory (not in
  `.claude/`). Output only valid TikZ code, with no markdown formatting or
  explanations
- Add this comment at the top of the file, saying how it was generated:
  ```latex
  % Generated by `.claude/skills/figure.create_tikz/SKILL.md`
  ```

## Render to Image
- Generate a PNG with the rendering script:
  ```bash
  > ./helpers_root/dev_scripts_helpers/dockerize/dockerized_tikz_to_bitmap.py \
      -i tikz_figure.tex \
      -o output.png
  ```

- Open the image to inspect it:
  ```bash
  > open output.png
  ```

## Iterate and Refine
- If the PNG differs from the original in significant ways:
  - Identify layout discrepancies
  - Update `./tikz_figure.tex` to match the input better
  - Re-render and verify

## Compare for User
- Build a side-by-side comparison and open it:
  ```bash
  > montage <ORIGINAL_IMAGE> output.png -tile 2x1 -geometry 500x+10+10 \
      -background white comparison.png
  > open comparison.png
  ```
- Remove `comparison.png` after the user has seen it. It is a temp file, not a
  deliverable

# Verification
- [ ] All lines connect cleanly to box edges with no overlap
- [ ] The diagram closely matches the original (same layout, colors, semantics)
- [ ] The TikZ file says that it was generated by
      `.claude/skills/figure.create_tikz/SKILL.md`

# Examples
- **Bad**: Converting a photograph of a natural scene into TikZ (infeasible; use
  image inclusion instead)
- **Good**: Converting a circuit diagram sketch into TikZ with accurate node
  positioning, labeled connections, and proper symmetry
receiving-code-reviewobra/superpowersUse when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementationStars: 322kUpdated Claude CodeCodexVerifiedInstall
requesting-code-reviewobra/superpowersUse when completing tasks, implementing major features, or before merging to verify work meets requirementsStars: 322kUpdated Claude CodeCodexVerifiedInstall
using-git-worktreesobra/superpowersUse when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallbackStars: 322kUpdated Claude CodeCodexVerifiedInstall
verification-before-completionobra/superpowersUse when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions alwaysStars: 322kUpdated Claude CodeCodexVerifiedInstall