paper-figure-artifact-audit

星标7

分支2

更新时间2026年6月20日 18:38

Audit paper/artifact figures and tables after final results are generated, focusing on visual clarity, caption consistency, numeric provenance, and submission-readiness. Use when the user says "check figures", "visualization audit", "图挤了", "caption/table audit", "artifact figure QA", or "投稿前检查图表".

安装

用 Codex 或 Claude 帮你安装复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。

来源

Yulivu

Yulivu/debuffer-skills

打开 GitHub 仓库查看创作者相关仓库

下载

Paper Figure Artifact Audit

Use this skill after final results are generated and before paper submission, artifact packaging, or Overleaf upload.

Workflow

Enumerate visuals and tables:
- PDF/PNG/SVG figures.
- CSV, Markdown, TeX, and notebook-generated tables.
- plotting scripts and trace CSV/provenance files.
Render PDF figures to PNG previews when visual inspection is needed.
Inspect visual clarity:
- axes, tick labels, legends, colorbars, annotations, panel labels.
- clipped text and PDF bounding boxes.
- crowded captions or captions pretending to be x-axis labels.
Verify numeric provenance:
- table values match current trace CSVs or promoted result files.
- figure curves/bars are generated from current final results.
- no old package data is silently reused.
Check captions:
- caption describes what the figure/table actually measures.
- caption does not strengthen the claim beyond the data.
- metric names are consistent across paper, artifact, and trace files.
Distinguish metrics carefully:
- latency speedup.
- storage ratio.
- fidelity/quality.
- recall.
- break-even or amortization metrics.
Fix visuals by editing plotting code/specs when possible, then regenerate. Avoid manual edits to generated PDFs.
Re-render previews and rerun checks before marking ready.

Visual Checks

Legends must not cover data lines.
Colorbars need labels separate from axes.
X-axis labels must not look like panel captions.
Multiline tick labels must remain readable.
Heatmap annotation text must contrast with cell colors.
PDF bounding boxes must not clip labels, legends, or captions.
Any text inside figures should match the paper body font size unless the venue requires a documented exception.

Guardrails

Do not alter numeric results during visual fixes.
Do not change captions to stronger claims than the data supports.
Public artifact visuals must stay clean and English.
Internal writing visuals can match public visuals, but READMEs may differ.
If provenance cannot be established, mark the item as blocked instead of approving it.

同仓库更多 Skills

同仓库

idea-discovery

Yulivu/debuffer-skills

Workflow 1: Full idea discovery pipeline. Orchestrates research-lit → idea-creator → novelty-check → research-review → research-refine plus experiment-plan to go from a broad research direction to validated, pilot-tested ideas. Also handles robotics / embodied AI directions through simulation-first robotics mode. Use when user says "找idea全流程", "idea discovery pipeline", "从零开始找方向", "robotics idea discovery", "embodied AI idea", or wants the complete idea exploration workflow.

2026-07-257

idea-creator

Yulivu/debuffer-skills

Generate and rank research ideas given a broad direction. Use when user says "找idea", "brainstorm ideas", "generate research ideas", "what can we work on", or wants to explore a research area for publishable directions.

2026-07-257

reference-paper-deconstruction

Yulivu/debuffer-skills

Turn one core reference paper and up to three comparable papers into evidence-grounded topic candidates. Use when the user asks to 拆解对标论文, 从论文找选题, analyze reference papers for research ideas, map conflicting findings, or wants a reference-led idea start before novelty checking.

2026-07-257

research-refine

Yulivu/debuffer-skills

Turn a vague research direction into a problem-anchored, elegant, frontier-aware, implementation-oriented method plan via iterative GPT-5.4 review. Use when the user says "refine my approach", "帮我细化方案", "decompose this problem", "打磨idea", "refine research plan", "细化研究方案", or wants a concrete research method that stays simple, focused, and top-venue ready instead of a vague or overbuilt idea.

2026-07-257

paper-plan

Yulivu/debuffer-skills

Generate a structured paper outline from audited formal experiment results and review conclusions. Use when user says "写大纲", "paper outline", "plan the paper", "论文规划", or wants a paper plan after formal runs and evidence audit. If only smoke, pilot, toy, or validation results exist, produce a gap report or next actions instead of routing to manuscript drafting.

2026-07-257

paper-write

Yulivu/debuffer-skills

Draft LaTeX paper section by section from an outline. Use when user says "写论文", "write paper", "draft LaTeX", "开始写", or wants to generate LaTeX content from a paper plan.

2026-07-257

name	paper-figure-artifact-audit
description	Audit paper/artifact figures and tables after final results are generated, focusing on visual clarity, caption consistency, numeric provenance, and submission-readiness. Use when the user says "check figures", "visualization audit", "图挤了", "caption/table audit", "artifact figure QA", or "投稿前检查图表".

Paper Figure Artifact Audit

Use this skill after final results are generated and before paper submission, artifact packaging, or Overleaf upload.

Workflow

Enumerate visuals and tables:
- PDF/PNG/SVG figures.
- CSV, Markdown, TeX, and notebook-generated tables.
- plotting scripts and trace CSV/provenance files.
Render PDF figures to PNG previews when visual inspection is needed.
Inspect visual clarity:
- axes, tick labels, legends, colorbars, annotations, panel labels.
- clipped text and PDF bounding boxes.
- crowded captions or captions pretending to be x-axis labels.
Verify numeric provenance:
- table values match current trace CSVs or promoted result files.
- figure curves/bars are generated from current final results.
- no old package data is silently reused.
Check captions:
- caption describes what the figure/table actually measures.
- caption does not strengthen the claim beyond the data.
- metric names are consistent across paper, artifact, and trace files.
Distinguish metrics carefully:
- latency speedup.
- storage ratio.
- fidelity/quality.
- recall.
- break-even or amortization metrics.
Fix visuals by editing plotting code/specs when possible, then regenerate. Avoid manual edits to generated PDFs.
Re-render previews and rerun checks before marking ready.

Visual Checks

Legends must not cover data lines.
Colorbars need labels separate from axes.
X-axis labels must not look like panel captions.
Multiline tick labels must remain readable.
Heatmap annotation text must contrast with cell colors.
PDF bounding boxes must not clip labels, legends, or captions.
Any text inside figures should match the paper body font size unless the venue requires a documented exception.

Guardrails

Do not alter numeric results during visual fixes.
Do not change captions to stronger claims than the data supports.
Public artifact visuals must stay clean and English.
Internal writing visuals can match public visuals, but READMEs may differ.
If provenance cannot be established, mark the item as blocked instead of approving it.