SkillAgentSearch skills...

radiology-figure

Create/audit publication plots, visual abstracts and imaging panels with render QA; not tables or decks. CN: 统计图、ROC、森林图、影像图版

Install / Use

npx skills add huang-sir1/radiology-skills --skill radiology-figure

Installs into whichever agent you are using.

About this skill
📄

SKILL.md

Installable skill definition

Quality Score

85/100

Supported Platforms

Universal

Our assessment of radiology-figure

radiology-figure scores 85/100 on our quality scale, 2618th of 4,700 Development & Engineering skills we index.

Its SKILL.md is 10 KB long, well organised into 8 sections and no code examples: a thorough specification that gives an agent plenty to work with.

With 1,785 GitHub stars, it is one of the more widely adopted skills in the catalogue.

Substance
29/30
Structure
13/20
Description
15/15
Adoption
14/20
Freshness
15/15

Maintenance, license and trust

  • The repository was last updated about 2 months ago, so radiology-figure is actively maintained.
  • Our last check on 2026-11-04 found the source still online.
  • No license is declared. By default that means all rights are reserved: you can read it, but reusing or redistributing it is not clearly permitted. Ask the author before building on it commercially.
  • Its trust signals score 88/100, with 1 caution from licensing, adoption, age or documentation. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.

Safety scan

No issues found

Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful.

AI review by kimi-k2.7-code on 2026-10-02. Automated pattern scan on 2026-10-02. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.

radiology-figure compared with similar skills

All 4 of these similar skills score higher than radiology-figure; compare them before choosing.

SkillScoreStarsUpdatedFormat
radiology-figure (this skill)by huang-sir1851.8k56d agoSKILL.md
taste-skillby Leonxlnx100116.8k8d agoCLAUDE.md
ai-job-searchby ai-job-search-dev10072.1ktodayCLAUDE.md
claude-howtoby luongnv8910047.3k2d agoCLAUDE.md
lobehubby lobehub10085.6ktodayMCP Server

Frequently asked questions

How do I install radiology-figure?
Run npx skills add huang-sir1/radiology-skills --skill radiology-figure. The install tabs above show the steps for each supported agent.
Which AI agents does radiology-figure work with?
It is written for Universal, as a SKILL.md file. Other agents that read the same format can often use it too.
Is radiology-figure safe to use?
Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful. It declares no license and scores 88/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
Is radiology-figure still maintained?
The repository was last updated about 2 months ago, so radiology-figure is actively maintained.

name: radiology-figure description: "Create/audit publication plots, visual abstracts and imaging panels with render QA; not tables or decks. CN: 统计图、ROC、森林图、影像图版"

Radiology Publication Figures

Use this skill to build figures that pass Radiology's technical and editorial bar: correct file format and resolution, legible typography, color-blind-safe palettes, honest axes, and the specific chart types imaging-AI reviewers expect (ROC, calibration, decision-curve, forest/SROC, Kaplan-Meier, Bland-Altman), plus de-identified annotated imaging panels.

Core stance

  • Vector first. Primary output is editable .svg (or .pdf); secondary is a ≥ 300 dpi raster (TIFF/PNG). Keep text as text (svg.fonttype='none'), not outlines, so editors can re-typeset.
  • One figure, one message. Each panel answers one question; no two panels duplicate it. Panels are labelled A, B, C (Radiology-family) or a, b, c (Nature-family — the case is venue-dependent, never mixed within one manuscript; see "When to open extra files").
  • Honest graphics. Axes start where the data demand (don't truncate to exaggerate); show uncertainty (CI bands, error bars); state n.
  • De-identify every image. No PHI burned into pixels, no faces/identifiers; scrub DICOM overlays; report windowing (WL/WW) and add a scale bar where size matters.
  • Match the journal. Sans-serif (Arial/Helvetica), figure width to column — Radiology-family single ~85 mm / double ~170 mm, or Nature-family single 89 mm / double 183 mm (max height 170 mm) — adequate font size at final print size (≈ 7–9 pt min). Confirm the target venue before sizing the first figure.
  • Never fabricate data. Plot only supplied/loaded values; mark simulated/example data clearly.

When to use

  • Statistical figures: ROC (+ DeLong annotation), calibration, decision-curve, forest, SROC, Kaplan-Meier (with numbers-at-risk), Bland-Altman, box/violin, heatmaps/clustermaps.
  • Radiogenomics: MOFA/factor plots, deconvolution stacked bars, habitat maps, correlation heatmaps.
  • Imaging panels: multi-row montages, before/after, arrows/insets, windowing labels, scale bars.
  • Flow diagrams: CONSORT / STARD / PRISMA patient-selection diagrams.

When to open extra files

| File | Open when | |---|---| | references/radiology-figure-guidelines.md | File format, resolution, size, fonts, color, panel labelling, de-identification rules | | references/chart-types.md | Choosing/parameterising the right statistical chart (ROC, calibration, DCA, forest, KM, Bland-Altman, heatmap) | | references/imaging-panels.md | Building montages: windowing, arrows, insets, scale bars, anonymisation, panel layout | | references/api.md | The matplotlib rcParams preamble, color palette, and reusable helper functions (ROC/calibration/forest/KM) | | references/design-theory.md | Typography, layout grid, color-blind-safe palettes, anti-redundancy, accessibility | | references/color-systems.md | Picking ONE palette (Okabe-Ito / NPG / Morandi) and mapping color→meaning so every figure matches | | references/survival-figures.md | Kaplan-Meier integrity (curve ↔ numbers-at-risk ↔ censoring), numbers-at-risk done right, time-dependent (IPCW) ROC/calibration/DCA, incremental-value framing | | references/figure-set-consistency.md | Unifying palette/fonts/axes across all figures, and cross-validating every figure number against the manuscript tables and data before export | | references/nature-figure-spec.md | Target is a Nature-portfolio venue instead of Radiology — column widths (89/183 mm), lowercase panel letters, RGB, legend word cap, Extended Data/Source Data display-item split | | references/figure-intent-and-render-qa.md | Full figure set planning, crowded/colliding labels, DCA/KM/heatmap layout problems, final-size render review, source-data crosswalk, or premium academic visual polish | | references/journal-family-visual-style.md | Target journal family is known, the user supplied author-guide PDFs/classic articles, or the figure set needs Nature/npj or European Radiology visual taste | | references/backend-and-geometry-qa.md | Choosing Python vs R, using survival/nomogram/ComplexHeatmap conventions, or auditing white background, excessive whitespace, edge clipping, and export geometry |

Workflow

Choose the requested scope first: advisory explains/designs a figure, audit inspects existing figures and reports findings, and create-edit produces or changes artifacts. Audit may render a temporary preview for inspection; it does not authorize rewriting originals. Generate scripts and exports only for create-edit or an explicitly requested artifact. Use only applicable QA items: conceptual diagrams and flowcharts need not have numerical axes, confidence intervals or sample n.

  1. Confirm the target venue (Radiology-family default, or Nature-family → nature-figure-spec.md) before sizing the first figure — column widths and panel-letter case differ and are painful to change after the set is built.
  2. For venue-specific visual taste, open journal-family-visual-style.md and apply the target family's panel lettering, legend density, graphical abstract, table, and source-data conventions.
  3. For full figure sets or layout-sensitive figures, open figure-intent-and-render-qa.md and create the figure intent table plus source-data crosswalk before drawing.
  4. Pick the chart for the message (chart-types.md). Discrimination → ROC; reliability → calibration; net benefit at justified risk thresholds linked to a clinical action → decision-curve; agreement → Bland-Altman; time-to-event → Kaplan-Meier; meta-analysis → forest/SROC; whole-study summary → graphical abstract.
  5. Resolve the backend with backend-and-geometry-qa.md: reuse the project's existing Python/R stack; otherwise default to Python for general imaging panels and consider R for survival/nomogram/ComplexHeatmap workflows where it is the established project route.
  6. Start the script with the backend theme and palette from api.md (Python) or backend-and-geometry-qa.md (R).
  7. Build the panel(s) with helper functions; add CI bands, n, and clear axis labels with units; label panels via panel_letter()/add_panel_letter() (api.md) with the case set for the confirmed venue — never hardcode chr(65+i) per script.
  8. For imaging panels, confirm de-identification, add windowing labels + scale bar + arrows; keep grayscale unless color encodes data.
  9. Export .svg/.pdf (editable text where possible) and a 300–600 dpi raster on an explicit opaque white canvas; check legibility at final print width.
  10. QA (see contract): run scripts/audit_figure_export.py on raster exports, then visually inspect the final render for text/number/legend overlap and semantic errors before returning.

Output contract

For advisory/audit scope, return only the requested plan or evidence-linked findings and applicable QA status. The script and file requirements below apply to artifact creation/editing. A diagnostic render is not proof that statistical, de-identification or submission checks passed.

  1. Figure plan — what each panel shows and why; the chart type chosen.
  2. Figure intent / source-data crosswalk — for full figure sets or submission figures, show what claim each panel supports and where the data came from.
  3. Script — a single runnable .py or .R using the selected backend theme; data inputs clearly marked (real vs example).
  4. Files — editable figure.svg/figure.pdf + figure.png/.tiff at ≥ 300 dpi.
  5. QA notes — fonts embedded as text, color-blind check, axis honesty, n shown, de-identification confirmed for any image panel, final-size render checked for no overlap/clipping, and venue-family visual style checked when applicable.

QA checklist (run before returning)

  • Selected backend theme/preamble is present; output is editable .svg/.pdf with text retained where possible and a ≥ 300 dpi raster on an opaque white canvas.
  • Every axis labelled with units; legend present; panels labelled A/B/C or a/b/c per the confirmed venue, consistently across the whole figure set.
  • Uncertainty shown (CI band/error bars) and n stated.
  • Color-blind-safe; not reliant on red/green alone; sufficient contrast.
  • Imaging panels de-identified; windowing + scale bar present where relevant.
  • No invented data points; example data flagged.
  • Final exported render inspected at target size; no label/tick/legend/number overlap, clipping, or text crossing plot elements.
  • Venue-family rules satisfied when applicable (panel-letter case, background, legend length, graphical abstract blocks, Source Data/table expectations).
  • Raster geometry audit recorded (white canvas, DPI metadata when available, occupied area, edge-ink/clipping risk); visual inspection still required for actual text collisions.

Handoffs

  • The statistic behind the plot (AUC CI, DeLong, ICC, calibration metrics, net benefit) → radiology-stats.
  • Whether the figure satisfies a checklist item (flow diagram for STARD/CONSORT), or the Reporting Summary/Nature Portfolio checklist → radiology-reporting.
  • Figure legends/captions prose, display-item plan (main vs Extended Data) → radiology-writing.
  • Source Data files, Extended Data vs Supplementary Information wording → radiology-data.
  • Reuse or adapt a frozen figure inside an audience-specific scientific deck → radiology-paper2ppt with the editable asset, source locator and transformation limits.
  • Full figure set finished and ready for a harsh read before submission → radiology-prereview.
  • Figure/table value reconciliation and project state → radiology-table / radiology-pipeline; upload package → radiology-submission.

Related Skills

View on GitHub
GitHub Stars1.8k
CategoryDevelopment
Updated1mo ago
Forks17

Languages

Python

Trust signals

88/100

From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.

1 medium