[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-80873":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":9,"language":10,"languages":9,"totalLinesOfCode":9,"stars":11,"forks":12,"watchers":13,"openIssues":14,"contributorsCount":14,"subscribersCount":14,"size":14,"stars1d":12,"stars7d":15,"stars30d":16,"stars90d":14,"forks30d":14,"starsTrendScore":17,"compositeScore":18,"rankGlobal":9,"rankLanguage":9,"license":19,"archived":20,"fork":20,"defaultBranch":21,"hasWiki":22,"hasPages":20,"topics":23,"createdAt":9,"pushedAt":9,"updatedAt":24,"readmeContent":25,"aiSummary":26,"trendingCount":14,"starSnapshotCount":14,"syncStatus":12,"lastSyncTime":27,"discoverSource":28},80873,"cvpr-2026-poster-skill","yunyiliu\u002Fcvpr-2026-poster-skill","yunyiliu","Reusable CVPR 2026 poster planning skill for Codex and Claude Code.",null,"Python",40,2,35,0,4,5,6,1.43,"MIT License",false,"main",true,[],"2026-06-12 02:04:07","# CVPR 2026 Poster Skill\n\nTurn a CVPR 2026 paper into a print-ready 84\"×42\" poster by chatting\nwith an AI agent. No manual templating, no PowerPoint, no fighting\nwith print dialogs.\n\nThis is a Claude Code \u002F Codex **skill** — not a standalone CLI tool.\nYou install it once into your agent, then talk to it in plain language.\n\n---\n\n## How to start\n\n### 1. Install the skill (one time)\n\n**Claude Code**\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fyunyiliu\u002Fcvpr-2026-poster-skill.git\nmkdir -p ~\u002F.claude\u002Fskills\nln -s \"$(pwd)\u002Fcvpr-2026-poster-skill\u002Fcvpr-2026-poster\" ~\u002F.claude\u002Fskills\u002Fcvpr-2026-poster\n```\n\nStart a new Claude Code session (existing sessions don't auto-pick-up\nnew skills). The skill appears in the available-skills list as\n`cvpr-2026-poster`.\n\n**Codex \u002F other agents**: copy or symlink the same `cvpr-2026-poster\u002F`\nfolder into your agent's skill location — refer to that tool's docs\nfor the exact path.\n\n### 2. Ask the agent to build the poster\n\nIn a new session, write a single message like:\n\n> Use `$cvpr-2026-poster` to create a poster workspace from my Overleaf\n> folder at `\u002Fpath\u002Fto\u002Fyour\u002Foverleaf-folder\u002F`. Fill the brief from\n> LaTeX, copy figures, auto-fetch institution logos, and sync the\n> editable poster.\n\nThe agent will:\n\n- Run `init_poster_project.py` to scaffold the workspace\n- Run `fill_brief_from_latex.py` to extract your paper's title,\n  authors, affiliations, abstract, and figures\n- Run `sync_poster_from_brief.py` to populate the editable HTML, copy\n  figures into place, and fetch logos from your institutions' websites\n- Open the resulting `poster\u002Findex.html` for you to inspect\n\nIf you have multiple institutions, list them in the same order as\n`Affiliations` in `poster_brief.md` and add one institution website\nper school. The agent does this automatically when the LaTeX is clear.\n\n### 3. Iterate by talking to the agent\n\nOnce the poster is open in the browser, just describe edits in chat:\n\n> Make the Results card use a side-by-side figure\u002Ftext layout, and\n> shrink the conference logo a bit.\n\n> The Contributions card is too long — keep it to one lead sentence\n> and three bullets.\n\n> Move the Conclusion card to the bottom of the rightmost column.\n\nThe agent will edit `poster-config.json` (and re-embed it into\n`index.html`) for you. Hard-refresh the browser tab to see the result.\n\nYou can also edit in the browser directly:\n\n- Drag column dividers to change column widths\n- Drag horizontal dividers to change card heights\n- Drag a card's `◆` handle (or click two handles in sequence) to\n  swap card positions\n- Use `A+` \u002F `A-` to scale all fonts globally\n- Click **Save** when you're happy — that downloads\n  `poster-config.json` to `~\u002FDownloads\u002F` with your layout\n\n### 4. Export to a print-ready PDF\n\nAfter you click **Save** in the browser, in a terminal:\n\n```bash\ncd poster-workspace\nbash bake_and_export.sh\n```\n\nThis picks up the saved `poster-config.json` from `~\u002FDownloads\u002F`,\nre-embeds it into `index.html`, and runs headless Chrome to produce\n`poster.pdf` at exactly **84.01\" × 42.01\"** (or the workshop size).\n\nIf you have not touched the layout in the browser since the last\nagent edit, just run:\n\n```bash\nbash export_pdf.sh\n```\n\nSend `poster.pdf` directly to your printer.\n\n---\n\n## What you need\n\n- Google Chrome (for the PDF export step)\n- Python 3 (preinstalled on macOS and most Linux distros)\n\nThat's all. The skill ships the rest:\n\n- LaTeX extractor\n- Institution-logo auto-fetcher\n- Editable HTML poster scaffold\n- Headless-Chrome PDF exporter at the exact poster size\n- Official CVPR 2026 Denver logo and template\n\n---\n\n## What gets created\n\nA typical workspace looks like:\n\n```\nposter-workspace\u002F\n├── poster_brief.md         ← human-readable summary of paper content\n├── poster\u002F\n│   ├── index.html          ← the editable poster, open in a browser\n│   ├── poster-config.json  ← serialized layout + content\n│   ├── figures\u002F            ← display copies of figures\n│   └── logos\u002F\n│       ├── user\u002F           ← school logos (top-left)\n│       └── official\u002F       ← CVPR conference logo (top-right)\n├── assets\u002F\n│   ├── figures\u002F            ← original figures from the LaTeX project\n│   └── logos\u002F              ← school \u002F lab logos (user or auto-fetched)\n├── references\u002F\n│   └── latex-extract.md    ← summary of what was pulled from LaTeX\n├── export_pdf.sh           ← one-command headless-Chrome PDF export\n└── bake_and_export.sh      ← bake current browser layout, then export\n```\n\n---\n\n## CVPR 2026 size reference\n\n| Track | Poster size | Aspect ratio |\n|-------|-------------|--------------|\n| Main \u002F Findings | 84\" × 42\" (2134 × 1067 mm) | 2 : 1 landscape |\n| Workshop | 42\" × 21\" (1067 × 533 mm) | 2 : 1 landscape |\n\nOutput is PDF, no bleed. The exporter honors the poster's `@page`\ndeclaration so the PDF page size matches the track exactly.\n\n---\n\n## Without an AI agent\n\nIf you want to drive the scripts manually instead of through an agent,\nall of them are plain Python:\n\n```bash\npython3 cvpr-2026-poster\u002Fscripts\u002Finit_poster_project.py --project-dir .\u002Fposter-workspace --track main\npython3 cvpr-2026-poster\u002Fscripts\u002Ffill_brief_from_latex.py --project-dir .\u002Fposter-workspace --latex-dir \u002Fpath\u002Fto\u002Foverleaf --copy-figures\n# edit poster-workspace\u002Fposter_brief.md\npython3 cvpr-2026-poster\u002Fscripts\u002Fsync_poster_from_brief.py --project-dir .\u002Fposter-workspace --fetch-logos-if-missing\nopen poster-workspace\u002Fposter\u002Findex.html\n# (edit in browser, click Save when done)\ncd poster-workspace && bash bake_and_export.sh\n```\n\n---\n\n## License\n\nMIT. See [LICENSE](LICENSE).\n\n","该项目提供了一种可复用的技能，用于将CVPR 2026会议论文转换成打印就绪的84\"×42\"海报，通过与AI代理对话实现，无需手动模板设置或使用PowerPoint。其核心功能包括自动提取论文标题、作者、机构等信息，并从指定路径同步图表和机构标志至自动生成的HTML文件中。此外，用户可通过自然语言指令调整海报布局和设计细节，如卡片位置、字体大小等。此技能适用于需要快速制作专业学术海报的研究人员或学生，特别是那些希望简化海报制作流程并提高效率的人士。","2026-06-11 04:02:37","CREATED_QUERY"]