Como usar
- Copie o comando e verifique o pacote antes de instalar no seu ambiente de agente.
- Consulte no original as ferramentas, credenciais e ambientes necessários.
- Descreva os dados de entrada e a entrega esperada; confira a saída com as fontes.
npx skills add https://github.com/suzheng/k2f --skill k2f-research-posterAntes de começar
- A instalação não inclui assinatura de modelo, franquia API ou acesso aos seus arquivos.
- Estrelas e instalações indicam atividade, não um teste independente de qualidade ou segurança.
Documento original do Skill
Renderizado do repositório de origem, preservando títulos, exemplos, código, tabelas, links e imagens.
K2F Research Poster — AI-Native Aesthetic Design
Purpose
Convert scientific manuscripts, experiment data, and academic notes into print-ready conference posters. Posters fail when treated as oversized papers; attendees scan from 1.5–2 meters away and spend under 90 seconds. This skill enforces scientific distillation, viewing-distance hierarchy, strict 1-page geometry, and high visual density while eliminating multi-page overflow and blank lower thirds.
Built on K2F
This skill produces conference posters as .K2F packages. Follow the k2f skill for authoring: unpack or init a workspace, edit the semantic JSON tree (content/root.json), customize styles/theme.json, pack/verify, and export. K2F separates semantic structure from styling, compiling an immutable render lock that guarantees pixel-identical output across PDF, PPTX, DOCX, and IDML.
Before you build, install the core skill: npx skills add suzheng/k2f --skill k2f.
When to Use
- Creating conference posters from paper drafts, preprints, or notes.
- Designing large-format scientific presentations (Arch E 48"×36", 36"×48", ISO A0/A1).
- Transforming multi-page manuscripts into a single-page visual synthesis.
- Exporting vector PDF, editable PPTX, or IDML.
Inputs
- Dimensions: Landscape (48"×36" / ISO A0) or portrait (36"×48" / A0). Default: 48"×36" landscape (3 columns).
- Findings: Abstract, methods, key results, and conclusions.
- Visuals: Figures, plots, diagrams, and logos.
- Metadata: Title, authors, affiliations, paper/repo URL.
Workflow
1. Understand
Distill content into three tiers:
- 3-Second: Title, logos, takeaway banner.
- 30-Second: Conclusion-driven section headers, key charts, metric callouts.
- 3-Minute: Concise methods, data tables, references.
2. Plan
- Word Budget: 400–600 words (limit: 300–800). Posters are visual abstracts.
- Grid: Pinned header, 3-column body (2–3 for portrait), footer (
catalog/content/ex_poster_shell.json). - Visuals: Dedicate >= 40% area to diagrams, plots, cards, or tables.
3. Build
- Shell: Paper size in
manifest.json. Wrap root inpage_shellwith pinnedheight(page_height - margins) andbreak_inside: "avoid". - Typography (
styles/theme.jsonfor 1.5–2 m): Title 72–90 pt (bold), Headers (h1/h2) 36–44 pt, Subheads/Metrics (h3/stat_number) 28–34 pt, Body (body) 22–26 pt (>= 18 pt min), Captions 16–18 pt. - Components: Convert statistics to cards (
ex_card_bands.json) or tables (ex_table.json). Use{ "fr": 1 }growers to prevent gaps.
4. Validate
- Run
pack_verify.py --workspace <path> --render. - Single-page check:
pages=1. Pages > 1 is a hard defect. - Bottom-fill check: Baselines align above footer without voids (
PAGE_UNDERFILL). - Inspect rendered PNGs for legibility and clipping.
5. Revise
When overflowing or unbalanced:
- Cut non-essential text before shrinking fonts.
- Convert sentences into bullets or metric cards.
- Adjust column gap or card padding in
theme.json. - Reduce role font sizes globally in
theme.jsonby 1–2 pt.
Rules & Constraints
- Strict 1-Page Envelope: Posters are single-canvas presentations; spilling to page 2 disrupts display boards. Pin
page_shellheight to content box (page_height - margins) to guarantee single-page compilation. - Viewing Distance Physics: Reading distance is 1.5–2 m; body text below 20 pt is unreadable to standing attendees. Prune text rather than shrinking typography.
- Conclusion-Driven Headers: Scanning attendees decide whether to stop in seconds. State findings (e.g., "Model Cuts Latency by 42%") instead of generic labels ("Results").
- Styles in Theme Only: Define styling in
styles/theme.json. Keeping semantic nodes clean allows global adjustments and predictable compilation. - High Contrast: Conference halls have varied lighting; maintain WCAG AAA contrast using calm academic palettes with vibrant accents.
Common Failure Modes
| Failure Mode | Root Cause | Solution |
|---|---|---|
| Multi-Page Overflow (`pages > 1`) | Unpinned height or excess text. | Pin page_shell height; prune text under 700 words. |
| Empty Lower Canvas (`PAGE_UNDERFILL`) | Short text without growers. | Add { "fr": 1 } growers or enlarge visuals. |
| Wall of Text | Raw paper text pasted. | Condense to 3–4 bullets per card; use metric callouts. |
| Tiny Typography | Document fonts (< 18 pt) used. | Scale in theme.json: Title >= 72 pt, H1 >= 36 pt, Body >= 22 pt. |
| Lopsided Columns | Unequal column distribution. | Rebalance sections (Col 1: Problem; Col 2: Methods/Results; Col 3: Impact). |
Quality Checklist
- [ ] Geometry: Pinned
page_shellmatching content box; strictlypages=1. - [ ] Word Budget: 300–800 words total across poster.
- [ ] Hierarchy: Title readable from 5 m; headers from 2 m; body from 1.5 m.
- [ ] Visual Ratio: >= 40% surface area in figures, metrics, or tables.
- [ ] Column Balance: Baselines align evenly without voids.
- [ ] Export Fidelity: Clean vector PDF and editable PPTX shapes.
