Academic Figure Skills
Three skills for paper figures: fig-analyze reads material, fig-draw picks a style and renders, fig-pptx builds an editable PPT.
- 188 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The figure description and source excerpts go to your environment's image model.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/academic-figure-skillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The figure description and source excerpts go to your environment's image model.
This is third-party code. Review the repository files before installing.
What it does
Academic Figure Skills is a set of three independently installable skills for diagrams in scientific papers. `fig-analyze` studies a repository, draft or reference image and outputs structured architecture and figure-plan descriptions. `fig-draw` shows style options before the first render, writes a specification, generates the image and checks it against the source. `fig-pptx` runs only after the user agrees and turns a finished diagram into an editable presentation. A separate mode builds, diagnoses or revises a drawing prompt without generating an image. The render audit is bound to SHA-256 hashes of the image and spec and records a `pass`, `fail` or `unverified` status for every node and edge. The `validate_render_audit.py` script checks that the record is complete. Generation is done by the environment's built-in `image_gen` tool, for example in Codex.
Who it is for. Authors of papers and theses who need a method or architecture diagram in a consistent style without drawing it by hand in an editor.
Good fit when
- An architecture diagram for a paper must be drawn from a code repository.
- A figure from another publication serves as a composition reference without copying its content.
- A drawing prompt is diagnosed and revised without rendering.
- A finished diagram moves into slides as editable objects.
Not a fit when
- You need a data plot, not a diagram.
- The agent environment has no image generation tool.
- The figure must copy someone else's original exactly, text included.
Example request
Analyze this repository and draw an architecture diagram: show three styles first and render after I choose.Limitations
The audit checks record completeness and hash binding but cannot see the image or prove the reviewer's claims true. Quality depends on your environment's image model. The author's README is in Chinese, with English example commands. Image generation needs access to the `image_gen` tool in the agent environment; without it only analysis, specification and checking remain. The skill installs globally with `-g`, so it applies in every project until removed.
How to disable. Delete the `fig-analyze`, `fig-draw` and `fig-pptx` folders from the agent's skills directory or reverse the install through `npx skills`.
FAQ
How do I install all three skills?
Run `npx skills add Azhi-ss/academic-figure-skills -g --all`. The `-l` flag lists them without installing.
Can I skip the prompt review?
Yes. Say to draw directly or not to show the prompt. If no style or reference is named, `fig-draw` first offers a style preview.
What does fig-pptx do?
It builds an editable PPT with a clean background from a finished diagram. It starts only after your consent, asked once the figure is delivered.
Which styles are included?
Three core profiles are built in: modern technical vector, illustrated modular and pastel airy, plus a contrast style for ablations. The style is chosen before the first render.
Can I use a figure from another paper as a reference?
Yes, but only composition, stroke, zoning and emphasis are matched. The skill does not copy the original's text, topology or conclusions.
Discussion
No questions or reviews yet.
Related
SkillsA skill that draws paper figures as vectors: architecture, mechanism and overview figures exported to SVG, PDF and PPTX
Skill for publication figures: matplotlib charts for papers in one house style, with vector and raster export from sample scripts.
A skill that prepares and audits figures for Elsevier-style papers: size, fonts, lines, resolution and export from Python and R
A skill that turns a research paper into detailed English prompts for image generators to produce top-conference-quality figures
A Codex skill turns a paper or method into a visual blueprint, then draws an architecture figure with built-in image generation.
Three skills for academic writing: a paper schema for image generation, rewriting text to sound less machine-made, and drafting from scratch
README in short
A figure is built from a specification with a per-node and per-edge audit, not from a single unchecked prompt.