CommunityImagengithub.com

figure-plan

Plan biomedical figure storyboards and panel inventories, assemble requested manuscript figures as editable PowerPoint decks, or review an existing whole-figure layout. Use for planning article figures, laying out main and supplementary panels, and designing figure layouts and figure assembly. Planning-only requests finish with the plan. Route scientific-story selection to paper-framing, individual data plots to scientific-plotting and mechanism panels to scientific-visual-explainer.

¿Qué es figure-plan?

figure-plan is a Claude Code agent skill that plan biomedical figure storyboards and panel inventories, assemble requested manuscript figures as editable PowerPoint decks, or review an existing whole-figure layout. Use for planning article figures, laying out main and supplementary panels, and designing figure layouts and figure assembly. Planning-only requests finish with the plan. Route scientific-story selection to paper-framing, individual data plots to scientific-plotting and mechanism panels to scientific-visual-explainer.

Compatible con✓Claude Code~Codex CLI~Cursor
npx skills add https://github.com/mianaz/biomed-skills/tree/HEAD/skills/figure-plan

Installed? Explore more Imagen skills: steipete/songsee, affaan-m/frontend-design, mattpocock/improve-codebase-architecture · View all 6 →

Preguntar en tu IA favorita

Abre un nuevo chat con esta habilidad de agente ya precargada.

Documentación

Figure Plan

Read shared principles.

Deliver the requested figure plan, assembly or layout review. For a complete manuscript assembly, default to main-figures.pptx and supplemental-figures.pptx; for a specified figure or subset, deliver that subset. One slide holds one complete figure, with independent panel assets and separate editable letters.

Prepare

Read writing principles when developing the scientific story, and visual standards when designing, assembling or reviewing artwork. Reuse an existing paper-framing brief, claim IDs and overall figure order when present; otherwise develop the argument from the available findings.

For a planning-only request, deliver the story and panel inventory. Otherwise, for assembly, carry the requested figures through panel preparation, PowerPoint assembly, export and inspection. For an audit, inspect the supplied assembly and report or repair at the requested scope. Reuse established story and panel decisions for local layout work. Missing data can use labeled placeholders in a draft.

1. Establish the story and figure count

Read the available manuscript, figure legends and figure/data inventory. Write the central discovery in one sentence: what was found, what understanding changes, and why it matters. Build the figure sequence around the biological questions and their answers. Distinguish measured results from proposed explanations where that distinction changes a claim.

If a target journal is known, use submission-check to retrieve current official limits on main figures, Extended Data / supplemental items, column widths and typography. Save the source, date and rule. Without a journal, record provisional width assumptions and proceed.

Propose an explicit count of main and supplemental figures with a reason for each:

  • Read result placement and apply the removal test: if omitting an experiment changes the central claim or leaves its strongest alternative explanation unresolved, keep its decisive comparison in main or link the supporting figure explicitly.
  • Read literature patterns when choosing multi-panel pairing archetypes or aspect-ratio priors from the 500-panel CNS corpus.
  • Use the plan template to record each figure's claim, role, panel sequence, main/supplement decision and linked supporting figure.

Done: the plan states both counts, one argument per figure, and how the sequence supports the central discovery within known journal limits.

2. Map each panel to evidence and a display

Inspect existing plots/images and their source records. Reuse valid finished panels directly when they fit the claim and final placement. Produce new panels through their responsible skill (always exported without publication letters):

Keep each quantitative panel linked to its analysis cohort and full result record. For features repeated across main and supplemental panels, apply analysis consistency and reconcile each panel with the same primary feature definition or its explicitly named adaptation. Use the shared family-scope rules when selecting panels: moving or omitting a panel alone leaves its comparison family unchanged. Record result-driven selection; route unresolved inference to biomed-stats.

Give each panel a stable placement ID separate from its movable display letter. Group panels by the inference they support (design → readout, image ↔ quantification, matched comparison, orthogonal support, or alternative-explanation test) and specify reading order and required adjacencies.

Evidence statePowerPoint treatment
readyInsert the actual independent letter-free panel asset.
data-readyPlot the available data via the plotting skill, then insert the panel (or use a solid rectangle labeled "Data available — plot pending" for a layout-only pass).
missingUse a dashed neutral rectangle labeled "Data pending: <comparison/readout>; <chart type>".
needs-sourceUse a solid neutral rectangle labeled "Source/export needed" when the asset cannot be separated or used at required quality.

Done: every planned panel has a stable ID, source or named gap, chart/schematic type, evidence group and clear state.

3. Assemble figures in PowerPoint

Read canvas sizing to choose figure width and solve height from the evidence groups, and PowerPoint assembly and export when building or modifying the decks.

  • Deliver the requested decks with slide counts matching the requested figure inventory (no cover or section-divider slides).
  • Import each panel as its own independent PowerPoint object preserving its native aspect ratio; reflow rows instead of flattening or stretching panels.
  • Add every panel letter as a separate, editable PowerPoint text box in Arial.
  • Arrange evidence groups first, then member panels, using deliberate gutters and visual hierarchy (e.g., hero overview or schematic paired with compact narrow/tall comparisons). Never import a precomposed whole figure from R, Python, SVG or HTML as the slide's figure.

When running blinded benchmark comparisons or training a layout ranker, read blinded evaluation and trained ranking.

Done: the requested decks match the inventory, each figure has a recorded export width and height, and each panel, placeholder and letter is independently selectable in the saved PPTX.

4. Review at final viewing size and export

Read layout review and run the effective typography audit on the rendered slides.

  • Verify effective text size (> 10 pt target, 8 pt floor), scientific typography (complete gene italics, upright proteins/units), fixed palette.csv mappings, label clearance, internal blank space and outer/within-envelope whitespace.
  • Repair panel-internal defects in the plotting/schematic source and placement/letter defects in PowerPoint, then re-render and re-inspect the affected slide.
  • Export each slide via PowerPoint → PDF → lossless PNG/TIFF at the intended physical size (600 dpi default) and build the HTML contact sheet for handoff.

Done: files open cleanly, each figure has a rendered QC record, exported appearance matches the PPTX at final manuscript scale, and draft versus ready status is explicit.

Individual skills in this repo

This repo contains 1 individual skill — each has its own dedicated page.

Skills relacionados