[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7FDC5GH3AXQR808HSKNEETD":3,"skill-same:01M7FDC5GH3AXQR808HSKNEETD":46},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":6,"ownerHtmlUrl":29,"ownerBlogUrl":6,"ownerDescription":30,"updatedAt":31,"install":32,"resources":45},"01M7FDC5GH3AXQR808HSKNEETD","paper-figure",null,"Generate publication-quality figures and tables from experiment results. Use when user says \"画图\", \"作图\", \"generate figures\", \"paper figures\", or needs plots for a paper.","CZX-LBJ\u002Fdeepseek-science","CZX-LBJ","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F204594689?v=4","01M7FC59WQ9PFXT04MKGJDCFFT","czx","01M7FC59XB2GW0BKWN9JTDBQS7",1,"MIT","https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Faris\u002Fskills\u002Fskills-codex-claude-review\u002Fpaper-figure\u002FSKILL.md",[],[],"2026-11-16 10:05:34",false,"---\nname: \"paper-figure\"\ndescription: \"Generate publication-quality figures and tables from experiment results. Use when user says \\\"画图\\\", \\\"作图\\\", \\\"generate figures\\\", \\\"paper figures\\\", or needs plots for a paper.\"\n---\n\n> Override for Codex users who want **Claude Code**, not a second Codex agent, to act as the reviewer. Install this package **after** `skills\u002Fskills-codex\u002F*`.\n>\n> This reviewer is a different model family from the Codex executor. Every overlay trace\u002Faudit records:\n>\n> ```yaml\n> review_independence: cross-family\n> acceptance_status: accepted\n> ```\n\n# Paper Figure: Publication-Quality Plots from Experiment Data\n\nGenerate all figures and tables for a paper based on: **$ARGUMENTS**\n\n## Scope: What This Skill Can and Cannot Do\n\n| Category | Can auto-generate? | Examples |\n|----------|-------------------|----------|\n| **Data-driven plots** | ✅ Yes | Line plots (training curves), bar charts (method comparison), scatter plots, heatmaps, box\u002Fviolin plots |\n| **Comparison tables** | ✅ Yes | LaTeX tables comparing prior bounds, method features, ablation results |\n| **Multi-panel figures** | ✅ Yes | Subfigure grids combining multiple plots (e.g., 3×3 dataset × method) |\n| **Architecture\u002Fpipeline diagrams** | ❌ No — manual | Model architecture, data flow diagrams, system overviews. At best can generate a rough TikZ skeleton, but **expect to draw these yourself** using tools like draw.io, Figma, or TikZ |\n| **Generated image grids** | ❌ No — manual | Grids of generated samples (e.g., GAN\u002Fdiffusion outputs). These come from running your model, not from this skill |\n| **Photographs \u002F screenshots** | ❌ No — manual | Real-world images, UI screenshots, qualitative examples |\n\n**In practice:** For a typical ML paper, this skill handles ~60% of figures (all data plots + tables). The remaining ~40% (hero figure, architecture diagram, qualitative results) need to be created manually and placed in `figures\u002F` before running `\u002Fpaper-write`. The skill will detect these as \"existing figures\" and preserve them.\n\n## Constants\n\n- **STYLE = `publication`** — Visual style preset. Options: `publication` (default, clean for print), `poster` (larger fonts), `slide` (bold colors)\n- **DPI = 300** — Output resolution\n- **FORMAT = `pdf`** — Output format. Options: `pdf` (vector, best for LaTeX), `png` (raster fallback)\n- **COLOR_PALETTE = `tab10`** — Default matplotlib color cycle. Options: `tab10`, `Set2`, `colorblind` (deuteranopia-safe)\n- **FONT_SIZE = 10** — Base font size (matches typical conference body text)\n- **FIG_DIR = `figures\u002F`** — Output directory for generated figures\n- **REVIEWER_MODEL = `claude-review`** — Claude reviewer invoked through the local `claude-review` MCP bridge. Set `CLAUDE_REVIEW_MODEL` if you need a specific Claude model override.\n\n## Inputs\n\n1. **PAPER_PLAN.md** — figure plan table (from `\u002Fpaper-plan`)\n2. **Experiment data** — JSON files, CSV files, or screen logs in `figures\u002F` or project root\n3. **Existing figures** — any manually created figures to preserve\n\nIf no PAPER_PLAN.md exists, scan for data files and ask the user which figures to generate.\n\n## Workflow\n\n### Step 1: Read Figure Plan\n\nParse the Figure Plan table from PAPER_PLAN.md:\n\n```markdown\n| ID | Type | Description | Data Source | Priority |\n|----|------|-------------|-------------|----------|\n| Fig 1 | Architecture | ... | manual | HIGH |\n| Fig 2 | Line plot | ... | figures\u002Fexp.json | HIGH |\n```\n\nIdentify:\n- Which figures can be auto-generated from data\n- Which need manual creation (architecture diagrams, etc.)\n- Which are comparison tables (generate as LaTeX)\n\n### Step 2: Set Up Plotting Environment\n\nCreate a shared style configuration script:\n\n```python\n# paper_plot_style.py — shared across all figure scripts\nimport matplotlib.pyplot as plt\nimport matplotlib\nmatplotlib.rcParams.update({\n    'font.size': FONT_SIZE,\n    'font.family': 'serif',\n    'font.serif': ['Times New Roman', 'Times', 'DejaVu Serif'],\n    'axes.labelsize': FONT_SIZE,\n    'axes.titlesize': FONT_SIZE + 1,\n    'xtick.labelsize': FONT_SIZE - 1,\n    'ytick.labelsize': FONT_SIZE - 1,\n    'legend.fontsize': FONT_SIZE - 1,\n    'figure.dpi': DPI,\n    'savefig.dpi': DPI,\n    'savefig.bbox': 'tight',\n    'savefig.pad_inches': 0.05,\n    'axes.grid': False,\n    'axes.spines.top': False,\n    'axes.spines.right': False,\n    'text.usetex': False,  # set True if LaTeX is available\n    'mathtext.fontset': 'stix',\n})\n\n# Color palette\nCOLORS = plt.cm.tab10.colors  # or Set2, or colorblind-safe\n\ndef save_fig(fig, name, fmt=FORMAT):\n    \"\"\"Save figure to FIG_DIR with consistent naming.\"\"\"\n    fig.savefig(f'{FIG_DIR}\u002F{name}.{fmt}')\n    print(f'Saved: {FIG_DIR}\u002F{name}.{fmt}')\n```\n\n### Step 3: Auto-Select Figure Type\n\nUse this decision tree for data-driven figures (inspired by Imbad0202\u002Facademic-research-skills):\n\n| Data Pattern | Recommended Type | Size |\n|-------------|-----------------|------|\n| X=time\u002Fsteps, Y=metric | Line plot | 0.48\\textwidth |\n| Methods × 1 metric | Bar chart | 0.48\\textwidth |\n| Methods × multiple metrics | Grouped bar \u002F radar | 0.95\\textwidth |\n| Two continuous variables | Scatter plot | 0.48\\textwidth |\n| Matrix \u002F grid values | Heatmap | 0.48\\textwidth |\n| Distribution comparison | Box\u002Fviolin plot | 0.48\\textwidth |\n| Multi-dataset results | Multi-panel (subfigure) | 0.95\\textwidth |\n| Prior work comparison | LaTeX table | — |\n\n### Step 4: Generate Each Figure\n\nFor each figure in the plan, create a standalone Python script:\n\n**Line plots** (training curves, scaling):\n```python\n# gen_fig2_training_curves.py\nfrom paper_plot_style import *\nimport json\n\nwith open('figures\u002Fexp_results.json') as f:\n    data = json.load(f)\n\nfig, ax = plt.subplots(1, 1, figsize=(5, 3.5))\nax.plot(data['steps'], data['fac_loss'], label='Factorized', color=COLORS[0])\nax.plot(data['steps'], data['crf_loss'], label='CRF-LR', color=COLORS[1])\nax.set_xlabel('Training Steps')\nax.set_ylabel('Cross-Entropy Loss')\nax.legend(frameon=False)\nsave_fig(fig, 'fig2_training_curves')\n```\n\n**Bar charts** (comparison, ablation):\n```python\nfig, ax = plt.subplots(1, 1, figsize=(5, 3))\nmethods = ['Baseline', 'Method A', 'Method B', 'Ours']\nvalues = [82.3, 85.1, 86.7, 89.2]\nbars = ax.bar(methods, values, color=[COLORS[i] for i in range(len(methods))])\nax.set_ylabel('Accuracy (%)')\n# Add value labels on bars\nfor bar, val in zip(bars, values):\n    ax.text(bar.get_x() + bar.get_width()\u002F2, bar.get_height() + 0.3,\n            f'{val:.1f}', ha='center', va='bottom', fontsize=FONT_SIZE-1)\nsave_fig(fig, 'fig3_comparison')\n```\n\n**Comparison tables** (LaTeX, for theory papers):\n```latex\n\\begin{table}[t]\n\\centering\n\\caption{Comparison of estimation error bounds. $n$: sample size, $D$: ambient dim, $d$: latent dim, $K$: subspaces, $n_k$: modes.}\n\\label{tab:bounds}\n\\begin{tabular}{lccc}\n\\toprule\nMethod & Rate & Depends on $D$? & Multi-modal? \\\\\n\\midrule\n\\citet{MinimaxOkoAS23} & $n^{-s'\u002FD}$ & Yes (curse) & No \\\\\n\\citet{ScoreMatchingdistributionrecovery} & $n^{-2\u002Fd}$ & No & No \\\\\n\\textbf{Ours} & $\\sqrt{\\sum n_k d_k \u002F n}$ & No & Yes \\\\\n\\bottomrule\n\\end{tabular}\n\\end{table}\n```\n\n**Architecture\u002Fpipeline diagrams** (MANUAL — outside this skill's scope):\n- These require manual creation using draw.io, Figma, Keynote, or TikZ\n- This skill can generate a rough TikZ skeleton as a starting point, but **do not expect publication-quality results**\n- If the figure already exists in `figures\u002F`, preserve it and generate only the LaTeX `\\includegraphics` snippet\n- Flag as `[MANUAL]` in the figure plan and `latex_includes.tex`\n\n### Step 5: Run All Scripts\n\n```bash\n# Run all figure generation scripts\nfor script in gen_fig*.py; do\n    python \"$script\"\ndone\n```\n\nVerify all output files exist and are non-empty. Then **render-then-verify**:\nre-open each RENDERED PDF\u002FPNG (not the script) and self-check — no clipped\nlabels, no legend covering data, every number\u002Flabel readable at final print\nsize. This self-check happens BEFORE the Step 7 review, so the reviewer's\nbudget goes to substance, not to catching clipped axes.\n\n### Step 6: Generate LaTeX Include Snippets\n\nFor each figure, output the LaTeX code to include it:\n\n```latex\n% === Fig 2: Training Curves ===\n\\begin{figure}[t]\n    \\centering\n    \\includegraphics[width=0.48\\textwidth]{figures\u002Ffig2_training_curves.pdf}\n    \\caption{Training curves comparing factorized and CRF-LR denoising.}\n    \\label{fig:training_curves}\n\\end{figure}\n```\n\nSave all snippets to `figures\u002Flatex_includes.tex` for easy copy-paste into the paper.\n\n### Step 7: Figure Quality Review with REVIEWER_MODEL\n\nSend figure descriptions and captions to Claude for review:\n\n```\nmcp__claude-review__review_start:\n  prompt: |\n    Review these figure\u002Ftable plans for a [VENUE] submission.\n\n    For each figure:\n    1. Is the caption informative and self-contained?\n    2. Does the figure type match the data being shown?\n    3. Is the comparison fair and clear?\n    4. Any missing baselines or ablations?\n    5. Would a different visualization be more effective?\n\n    [list all figures with captions and descriptions]\n```\n\nAfter this start call, immediately save the returned `jobId` and poll `mcp__claude-review__review_status` with a bounded `waitSeconds` until `done=true`. Treat the completed status payload's `response` as the reviewer output, and save the completed `threadId` for any follow-up round.\n\n### Step 8: Quality Checklist\n\nThe checklist is PARTITIONED (pattern from Anthropic's Claude Science\n`figure-style` skill, Apache-2.0): **correctness rules always bind** — they are\nabout whether the figure tells the truth, have no aesthetic content, and no\nstyle choice may override them; **guidance rules are defaults** — they produce\na clean result, but a deliberate, stated alternative may override them.\n\n**Correctness — always binds, verify against the DATA before the render:**\n\n- [ ] **Excluded data never enters summaries** — a row excluded\u002Fflagged in the\n      source either disappears entirely or is drawn visibly distinct (open \u002F\n      hatched marker, named in the key); it never feeds a mean\u002FCI plotted\n      alongside included rows\n- [ ] **Captions and any claim-like title text are tested against EVERY plotted\n      row** — if one category contradicts the claim, qualify it (\"on 3 of 4\n      benchmarks\") or downgrade to a description; a figure that overclaims is\n      wrong even if it renders beautifully\n- [ ] **Comparable conditions only** — arms measured under different N \u002F budget\n      \u002F protocol are not drawn as visual peers; separate them or mark the\n      difference in the caption\n- [ ] **State n and what was held fixed** — every panel with a summary mark\n      says n and the unit of replication (panel or caption)\n- [ ] **Render-then-verify** — the Step-5 self-check on the RENDERED PDF\u002FPNG\n      (not the script) actually happened: no clipped labels, no legend covering\n      data, every number\u002Flabel readable at final print size\n\n**Guidance — strong defaults (from pedrohcgs\u002Fclaude-code-my-workflow), a\ndeliberate stated alternative may override — EXCEPT items that Key Rules below\nmake hard (vector-PDF output and no-titles-inside-figures are Key Rules: treat\nthose two as binding, not overridable):**\n\n- [ ] Font size readable at printed paper size (not too small)\n- [ ] Colors distinguishable in grayscale (print-friendly)\n- [ ] **No title inside figures** — titles go only in LaTeX `\\caption{}` (from pedrohcgs)\n- [ ] Legend does not overlap data\n- [ ] Axis labels have units where applicable\n- [ ] Axis labels are publication-quality (not variable names like `emp_rate`)\n- [ ] Figure width fits single column (0.48\\textwidth) or full width (0.95\\textwidth)\n- [ ] PDF output is vector (not rasterized text)\n- [ ] No matplotlib default title (remove `plt.title` for publications)\n- [ ] Serif font matches paper body text (Times \u002F Computer Modern)\n- [ ] Colorblind-accessible (if using colorblind palette)\n\n## Output\n\n```\nfigures\u002F\n├── paper_plot_style.py          # shared style config\n├── gen_fig1_architecture.py     # per-figure scripts\n├── gen_fig2_training_curves.py\n├── gen_fig3_comparison.py\n├── fig1_architecture.pdf        # generated figures\n├── fig2_training_curves.pdf\n├── fig3_comparison.pdf\n├── latex_includes.tex           # LaTeX snippets for all figures\n└── TABLE_*.tex                  # standalone table LaTeX files\n```\n\n## Key Rules\n\n- **Every figure must be reproducible** — save the generation script alongside the output\n- **Do NOT hardcode data** — always read from JSON\u002FCSV files\n- **Use vector format (PDF)** for all plots — PNG only as fallback\n- **No decorative elements** — no background colors, no 3D effects, no chart junk\n- **Consistent style across all figures** — same fonts, colors, line widths\n- **Colorblind-safe** — verify with https:\u002F\u002Fdavidmathlogic.com\u002Fcolorblind\u002F if needed\n- **One script per figure** — easy to re-run individual figures when data changes\n- **No titles inside figures** — captions are in LaTeX only\n- **Comparison tables count as figures** — generate them as standalone .tex files\n\n## Figure Type Reference\n\n| Type | When to Use | Typical Size |\n|------|------------|--------------|\n| Line plot | Training curves, scaling trends | 0.48\\textwidth |\n| Bar chart | Method comparison, ablation | 0.48\\textwidth |\n| Grouped bar | Multi-metric comparison | 0.95\\textwidth |\n| Scatter plot | Correlation analysis | 0.48\\textwidth |\n| Heatmap | Attention, confusion matrix | 0.48\\textwidth |\n| Box\u002Fviolin | Distribution comparison | 0.48\\textwidth |\n| Architecture | System overview | 0.95\\textwidth |\n| Multi-panel | Combined results (subfigures) | 0.95\\textwidth |\n| Comparison table | Prior bounds vs. ours (theory) | full width |\n\n## Acknowledgements\n\nDesign pattern (type × style matrix) inspired by [baoyu-skills](https:\u002F\u002Fgithub.com\u002Fjimliu\u002Fbaoyu-skills). Publication style defaults and figure rules from [pedrohcgs\u002Fclaude-code-my-workflow](https:\u002F\u002Fgithub.com\u002Fpedrohcgs\u002Fclaude-code-my-workflow). Visualization decision tree from [Imbad0202\u002Facademic-research-skills](https:\u002F\u002Fgithub.com\u002FImbad0202\u002Facademic-research-skills).\n","> Override for Codex users who want **Claude Code**, not a second Codex agent, to act as the reviewer. Install this package **after** `skills\u002Fskills-codex\u002F*`.\n>\n> This reviewer is a different model family from the Codex executor. Every overlay trace\u002Faudit records:\n>\n> ```yaml\n> review_independence: cross-family\n> acceptance_status: accepted\n> ```\n\n# Paper Figure: Publication-Quality Plots from Experiment Data\n\nGenerate all figures and tables for a paper based on: **$ARGUMENTS**\n\n## Scope: What This Skill Can and Cannot Do\n\n| Category | Can auto-generate? | Examples |\n|----------|-------------------|----------|\n| **Data-driven plots** | ✅ Yes | Line plots (training curves), bar charts (method comparison), scatter plots, heatmaps, box\u002Fviolin plots |\n| **Comparison tables** | ✅ Yes | LaTeX tables comparing prior bounds, method features, ablation results |\n| **Multi-panel figures** | ✅ Yes | Subfigure grids combining multiple plots (e.g., 3×3 dataset × method) |\n| **Architecture\u002Fpipeline diagrams** | ❌ No — manual | Model architecture, data flow diagrams, system overviews. At best can generate a rough TikZ skeleton, but **expect to draw these yourself** using tools like draw.io, Figma, or TikZ |\n| **Generated image grids** | ❌ No — manual | Grids of generated samples (e.g., GAN\u002Fdiffusion outputs). These come from running your model, not from this skill |\n| **Photographs \u002F screenshots** | ❌ No — manual | Real-world images, UI screenshots, qualitative examples |\n\n**In practice:** For a typical ML paper, this skill handles ~60% of figures (all data plots + tables). The remaining ~40% (hero figure, architecture diagram, qualitative results) need to be created manually and placed in `figures\u002F` before running `\u002Fpaper-write`. The skill will detect these as \"existing figures\" and preserve them.\n\n## Constants\n\n- **STYLE = `publication`** — Visual style preset. Options: `publication` (default, clean for print), `poster` (larger fonts), `slide` (bold colors)\n- **DPI = 300** — Output resolution\n- **FORMAT = `pdf`** — Output format. Options: `pdf` (vector, best for LaTeX), `png` (raster fallback)\n- **COLOR_PALETTE = `tab10`** — Default matplotlib color cycle. Options: `tab10`, `Set2`, `colorblind` (deuteranopia-safe)\n- **FONT_SIZE = 10** — Base font size (matches typical conference body text)\n- **FIG_DIR = `figures\u002F`** — Output directory for generated figures\n- **REVIEWER_MODEL = `claude-review`** — Claude reviewer invoked through the local `claude-review` MCP bridge. Set `CLAUDE_REVIEW_MODEL` if you need a specific Claude model override.\n\n## Inputs\n\n1. **PAPER_PLAN.md** — figure plan table (from `\u002Fpaper-plan`)\n2. **Experiment data** — JSON files, CSV files, or screen logs in `figures\u002F` or project root\n3. **Existing figures** — any manually created figures to preserve\n\nIf no PAPER_PLAN.md exists, scan for data files and ask the user which figures to generate.\n\n## Workflow\n\n### Step 1: Read Figure Plan\n\nParse the Figure Plan table from PAPER_PLAN.md:\n\n```markdown\n| ID | Type | Description | Data Source | Priority |\n|----|------|-------------|-------------|----------|\n| Fig 1 | Architecture | ... | manual | HIGH |\n| Fig 2 | Line plot | ... | figures\u002Fexp.json | HIGH |\n```\n\nIdentify:\n- Which figures can be auto-generated from data\n- Which need manual creation (architecture diagrams, etc.)\n- Which are comparison tables (generate as LaTeX)\n\n### Step 2: Set Up Plotting Environment\n\nCreate a shared style configuration script:\n\n```python\n# paper_plot_style.py — shared across all figure scripts\nimport matplotlib.pyplot as plt\nimport matplotlib\nmatplotlib.rcParams.update({\n    'font.size': FONT_SIZE,\n    'font.family': 'serif',\n    'font.serif': ['Times New Roman', 'Times', 'DejaVu Serif'],\n    'axes.labelsize': FONT_SIZE,\n    'axes.titlesize': FONT_SIZE + 1,\n    'xtick.labelsize': FONT_SIZE - 1,\n    'ytick.labelsize': FONT_SIZE - 1,\n    'legend.fontsize': FONT_SIZE - 1,\n    'figure.dpi': DPI,\n    'savefig.dpi': DPI,\n    'savefig.bbox': 'tight',\n    'savefig.pad_inches': 0.05,\n    'axes.grid': False,\n    'axes.spines.top': False,\n    'axes.spines.right': False,\n    'text.usetex': False,  # set True if LaTeX is available\n    'mathtext.fontset': 'stix',\n})\n\n# Color palette\nCOLORS = plt.cm.tab10.colors  # or Set2, or colorblind-safe\n\ndef save_fig(fig, name, fmt=FORMAT):\n    \"\"\"Save figure to FIG_DIR with consistent naming.\"\"\"\n    fig.savefig(f'{FIG_DIR}\u002F{name}.{fmt}')\n    print(f'Saved: {FIG_DIR}\u002F{name}.{fmt}')\n```\n\n### Step 3: Auto-Select Figure Type\n\nUse this decision tree for data-driven figures (inspired by Imbad0202\u002Facademic-research-skills):\n\n| Data Pattern | Recommended Type | Size |\n|-------------|-----------------|------|\n| X=time\u002Fsteps, Y=metric | Line plot | 0.48\\textwidth |\n| Methods × 1 metric | Bar chart | 0.48\\textwidth |\n| Methods × multiple metrics | Grouped bar \u002F radar | 0.95\\textwidth |\n| Two continuous variables | Scatter plot | 0.48\\textwidth |\n| Matrix \u002F grid values | Heatmap | 0.48\\textwidth |\n| Distribution comparison | Box\u002Fviolin plot | 0.48\\textwidth |\n| Multi-dataset results | Multi-panel (subfigure) | 0.95\\textwidth |\n| Prior work comparison | LaTeX table | — |\n\n### Step 4: Generate Each Figure\n\nFor each figure in the plan, create a standalone Python script:\n\n**Line plots** (training curves, scaling):\n```python\n# gen_fig2_training_curves.py\nfrom paper_plot_style import *\nimport json\n\nwith open('figures\u002Fexp_results.json') as f:\n    data = json.load(f)\n\nfig, ax = plt.subplots(1, 1, figsize=(5, 3.5))\nax.plot(data['steps'], data['fac_loss'], label='Factorized', color=COLORS[0])\nax.plot(data['steps'], data['crf_loss'], label='CRF-LR', color=COLORS[1])\nax.set_xlabel('Training Steps')\nax.set_ylabel('Cross-Entropy Loss')\nax.legend(frameon=False)\nsave_fig(fig, 'fig2_training_curves')\n```\n\n**Bar charts** (comparison, ablation):\n```python\nfig, ax = plt.subplots(1, 1, figsize=(5, 3))\nmethods = ['Baseline', 'Method A', 'Method B', 'Ours']\nvalues = [82.3, 85.1, 86.7, 89.2]\nbars = ax.bar(methods, values, color=[COLORS[i] for i in range(len(methods))])\nax.set_ylabel('Accuracy (%)')\n# Add value labels on bars\nfor bar, val in zip(bars, values):\n    ax.text(bar.get_x() + bar.get_width()\u002F2, bar.get_height() + 0.3,\n            f'{val:.1f}', ha='center', va='bottom', fontsize=FONT_SIZE-1)\nsave_fig(fig, 'fig3_comparison')\n```\n\n**Comparison tables** (LaTeX, for theory papers):\n```latex\n\\begin{table}[t]\n\\centering\n\\caption{Comparison of estimation error bounds. $n$: sample size, $D$: ambient dim, $d$: latent dim, $K$: subspaces, $n_k$: modes.}\n\\label{tab:bounds}\n\\begin{tabular}{lccc}\n\\toprule\nMethod & Rate & Depends on $D$? & Multi-modal? \\\\\n\\midrule\n\\citet{MinimaxOkoAS23} & $n^{-s'\u002FD}$ & Yes (curse) & No \\\\\n\\citet{ScoreMatchingdistributionrecovery} & $n^{-2\u002Fd}$ & No & No \\\\\n\\textbf{Ours} & $\\sqrt{\\sum n_k d_k \u002F n}$ & No & Yes \\\\\n\\bottomrule\n\\end{tabular}\n\\end{table}\n```\n\n**Architecture\u002Fpipeline diagrams** (MANUAL — outside this skill's scope):\n- These require manual creation using draw.io, Figma, Keynote, or TikZ\n- This skill can generate a rough TikZ skeleton as a starting point, but **do not expect publication-quality results**\n- If the figure already exists in `figures\u002F`, preserve it and generate only the LaTeX `\\includegraphics` snippet\n- Flag as `[MANUAL]` in the figure plan and `latex_includes.tex`\n\n### Step 5: Run All Scripts\n\n```bash\n# Run all figure generation scripts\nfor script in gen_fig*.py; do\n    python \"$script\"\ndone\n```\n\nVerify all output files exist and are non-empty. Then **render-then-verify**:\nre-open each RENDERED PDF\u002FPNG (not the script) and self-check — no clipped\nlabels, no legend covering data, every number\u002Flabel readable at final print\nsize. This self-check happens BEFORE the Step 7 review, so the reviewer's\nbudget goes to substance, not to catching clipped axes.\n\n### Step 6: Generate LaTeX Include Snippets\n\nFor each figure, output the LaTeX code to include it:\n\n```latex\n% === Fig 2: Training Curves ===\n\\begin{figure}[t]\n    \\centering\n    \\includegraphics[width=0.48\\textwidth]{figures\u002Ffig2_training_curves.pdf}\n    \\caption{Training curves comparing factorized and CRF-LR denoising.}\n    \\label{fig:training_curves}\n\\end{figure}\n```\n\nSave all snippets to `figures\u002Flatex_includes.tex` for easy copy-paste into the paper.\n\n### Step 7: Figure Quality Review with REVIEWER_MODEL\n\nSend figure descriptions and captions to Claude for review:\n\n```\nmcp__claude-review__review_start:\n  prompt: |\n    Review these figure\u002Ftable plans for a [VENUE] submission.\n\n    For each figure:\n    1. Is the caption informative and self-contained?\n    2. Does the figure type match the data being shown?\n    3. Is the comparison fair and clear?\n    4. Any missing baselines or ablations?\n    5. Would a different visualization be more effective?\n\n    [list all figures with captions and descriptions]\n```\n\nAfter this start call, immediately save the returned `jobId` and poll `mcp__claude-review__review_status` with a bounded `waitSeconds` until `done=true`. Treat the completed status payload's `response` as the reviewer output, and save the completed `threadId` for any follow-up round.\n\n### Step 8: Quality Checklist\n\nThe checklist is PARTITIONED (pattern from Anthropic's Claude Science\n`figure-style` skill, Apache-2.0): **correctness rules always bind** — they are\nabout whether the figure tells the truth, have no aesthetic content, and no\nstyle choice may override them; **guidance rules are defaults** — they produce\na clean result, but a deliberate, stated alternative may override them.\n\n**Correctness — always binds, verify against the DATA before the render:**\n\n- [ ] **Excluded data never enters summaries** — a row excluded\u002Fflagged in the\n      source either disappears entirely or is drawn visibly distinct (open \u002F\n      hatched marker, named in the key); it never feeds a mean\u002FCI plotted\n      alongside included rows\n- [ ] **Captions and any claim-like title text are tested against EVERY plotted\n      row** — if one category contradicts the claim, qualify it (\"on 3 of 4\n      benchmarks\") or downgrade to a description; a figure that overclaims is\n      wrong even if it renders beautifully\n- [ ] **Comparable conditions only** — arms measured under different N \u002F budget\n      \u002F protocol are not drawn as visual peers; separate them or mark the\n      difference in the caption\n- [ ] **State n and what was held fixed** — every panel with a summary mark\n      says n and the unit of replication (panel or caption)\n- [ ] **Render-then-verify** — the Step-5 self-check on the RENDERED PDF\u002FPNG\n      (not the script) actually happened: no clipped labels, no legend covering\n      data, every number\u002Flabel readable at final print size\n\n**Guidance — strong defaults (from pedrohcgs\u002Fclaude-code-my-workflow), a\ndeliberate stated alternative may override — EXCEPT items that Key Rules below\nmake hard (vector-PDF output and no-titles-inside-figures are Key Rules: treat\nthose two as binding, not overridable):**\n\n- [ ] Font size readable at printed paper size (not too small)\n- [ ] Colors distinguishable in grayscale (print-friendly)\n- [ ] **No title inside figures** — titles go only in LaTeX `\\caption{}` (from pedrohcgs)\n- [ ] Legend does not overlap data\n- [ ] Axis labels have units where applicable\n- [ ] Axis labels are publication-quality (not variable names like `emp_rate`)\n- [ ] Figure width fits single column (0.48\\textwidth) or full width (0.95\\textwidth)\n- [ ] PDF output is vector (not rasterized text)\n- [ ] No matplotlib default title (remove `plt.title` for publications)\n- [ ] Serif font matches paper body text (Times \u002F Computer Modern)\n- [ ] Colorblind-accessible (if using colorblind palette)\n\n## Output\n\n```\nfigures\u002F\n├── paper_plot_style.py          # shared style config\n├── gen_fig1_architecture.py     # per-figure scripts\n├── gen_fig2_training_curves.py\n├── gen_fig3_comparison.py\n├── fig1_architecture.pdf        # generated figures\n├── fig2_training_curves.pdf\n├── fig3_comparison.pdf\n├── latex_includes.tex           # LaTeX snippets for all figures\n└── TABLE_*.tex                  # standalone table LaTeX files\n```\n\n## Key Rules\n\n- **Every figure must be reproducible** — save the generation script alongside the output\n- **Do NOT hardcode data** — always read from JSON\u002FCSV files\n- **Use vector format (PDF)** for all plots — PNG only as fallback\n- **No decorative elements** — no background colors, no 3D effects, no chart junk\n- **Consistent style across all figures** — same fonts, colors, line widths\n- **Colorblind-safe** — verify with https:\u002F\u002Fdavidmathlogic.com\u002Fcolorblind\u002F if needed\n- **One script per figure** — easy to re-run individual figures when data changes\n- **No titles inside figures** — captions are in LaTeX only\n- **Comparison tables count as figures** — generate them as standalone .tex files\n\n## Figure Type Reference\n\n| Type | When to Use | Typical Size |\n|------|------------|--------------|\n| Line plot | Training curves, scaling trends | 0.48\\textwidth |\n| Bar chart | Method comparison, ablation | 0.48\\textwidth |\n| Grouped bar | Multi-metric comparison | 0.95\\textwidth |\n| Scatter plot | Correlation analysis | 0.48\\textwidth |\n| Heatmap | Attention, confusion matrix | 0.48\\textwidth |\n| Box\u002Fviolin | Distribution comparison | 0.48\\textwidth |\n| Architecture | System overview | 0.95\\textwidth |\n| Multi-panel | Combined results (subfigures) | 0.95\\textwidth |\n| Comparison table | Prior bounds vs. ours (theory) | full width |\n\n## Acknowledgements\n\nDesign pattern (type × style matrix) inspired by [baoyu-skills](https:\u002F\u002Fgithub.com\u002Fjimliu\u002Fbaoyu-skills). Publication style defaults and figure rules from [pedrohcgs\u002Fclaude-code-my-workflow](https:\u002F\u002Fgithub.com\u002Fpedrohcgs\u002Fclaude-code-my-workflow). Visualization decision tree from [Imbad0202\u002Facademic-research-skills](https:\u002F\u002Fgithub.com\u002FImbad0202\u002Facademic-research-skills).\n",{"name":5,"description":7},"main","18193d71ea3441d7c033dbf195b8a7c75ac96f75","packages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Faris\u002Fskills\u002Fskills-codex-claude-review\u002Fpaper-figure",0,"https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science","https:\u002F\u002Fgithub.com\u002FCZX-LBJ","一个没事喜欢随便写点代码的小菜鸡\r\n","2026-10-22 09:48:04",{"source":8,"skillName":5,"commands":33,"prompt":43,"zipAvailable":44},[34,37,40],{"tool":35,"command":36},"npx","npx skills add CZX-LBJ\u002Fdeepseek-science --skill paper-figure",{"tool":38,"command":39},"bunx","bunx skills add CZX-LBJ\u002Fdeepseek-science --skill paper-figure",{"tool":41,"command":42},"pnpm","pnpm dlx skills add CZX-LBJ\u002Fdeepseek-science --skill paper-figure","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Faris\u002Fskills\u002Fskills-codex-claude-review\u002Fpaper-figure\u002FSKILL.md",true,[],[47,54,61,68,75,82],{"skillNo":48,"name":49,"nameZh":6,"description":50,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":51,"categories":52,"keywords":53,"publishedAt":19},"01M7FEEKSNMRMKW6JEK93PFDFY","bio-free-energy-calculations","Performs alchemical free-energy calculations including relative binding free energy (RBFE \u002F FEP+) and absolute binding free energy (ABFE) via OpenFE, FEP+, GROMACS, AMBER pmemd, and OpenMM with explicit lambda scheduling, soft-core potentials, MBAR\u002FBAR analysis, cycle-closure validation, and protocol-appropriate enhanced sampling. Compares ML alternatives (Boltz-2 affinity, DeepDock). Use when ranking analogs by binding affinity beyond docking accuracy, performing prospective lead optimization, or validating SAR predictions.","https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Fbioskills\u002Fchemoinformatics\u002Ffree-energy-calculations\u002FSKILL.md",[],[],{"skillNo":55,"name":56,"nameZh":6,"description":57,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":58,"categories":59,"keywords":60,"publishedAt":19},"01M7FEE5CTE5QK55CEMXNNVWY9","bio-conformer-generation","Generates 3D conformer ensembles using RDKit ETKDGv3 with knowledge-enhanced distance geometry, MMFF94\u002FUFF force-field optimization, CREST + GFN2-xTB semi-empirical refinement, and macrocycle-aware torsion preferences. Provides explicit decision rules for single vs ensemble conformer use, RMSD pruning, energy windows, conformer count, and force-field choice. Use when preparing 3D ligands for docking, generating descriptor input for 3D QSAR, or sampling macrocycle\u002Fpeptide conformational ensembles.","https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Fbioskills\u002Fchemoinformatics\u002Fconformer-generation\u002FSKILL.md",[],[],{"skillNo":62,"name":63,"nameZh":6,"description":64,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":65,"categories":66,"keywords":67,"publishedAt":19},"01M7FEDNGFEMFWD1MSEDR5H7CY","bio-admet-prediction","Predicts ADMET properties using ADMETlab 3.0 (119 platform features, including 77 prediction models with modeled-endpoint uncertainty), ADMET-AI, DeepChem MolNet, and chemprop D-MPNN with explicit handling of OECD QSAR principles, applicability domain assessment, calibration, hERG\u002FCYP\u002FAMES endpoints, and PAINS \u002F Lipinski \u002F Ro5 \u002F Veber \u002F BBB druglikeness filters. Use when filtering compounds for drug-likeness, prioritizing leads by predicted safety, or building an in-house ADMET QSAR model.","https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Fbioskills\u002Fchemoinformatics\u002Fadmet-prediction\u002FSKILL.md",[],[],{"skillNo":69,"name":70,"nameZh":6,"description":71,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":72,"categories":73,"keywords":74,"publishedAt":19},"01M7FED69GFB54FRF7AV5066DN","writing-systems-papers","Paragraph-level structural blueprint for 10-12 page systems papers targeting OSDI, SOSP, ASPLOS, NSDI, and EuroSys. Provides page allocation, paragraph templates, and writing patterns. Use when user says \"写系统论文\", \"systems paper structure\", \"OSDI paper\", \"SOSP paper\", or wants fine-grained structural guidance for a systems conference submission.","https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Faris\u002Fskills\u002Fwriting-systems-papers\u002FSKILL.md",[],[],{"skillNo":76,"name":77,"nameZh":6,"description":78,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":79,"categories":80,"keywords":81,"publishedAt":19},"01M7FECVVDWZ0X0VEDTCE4B4DZ","wiki-enrich","Fill in the per-paper TODO sections of research-wiki\u002Fpapers\u002F\u003Cslug>.md pages that literature-ingest skills leave as bare scaffolds. Use when user says 'enrich wiki', 'fill paper TODOs', 'wiki body 補完', '把 paper 摘要寫進 wiki', 'research-wiki 自動填', or after a batch ingest that left papers\u002F as TODO scaffolds.","https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Faris\u002Fskills\u002Fwiki-enrich\u002FSKILL.md",[],[],{"skillNo":83,"name":84,"nameZh":6,"description":85,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":86,"categories":87,"keywords":88,"publishedAt":19},"01M7FECD4BCJ2MTSBDDSE3ZAJ5","web-debug-search","Search GitHub, Stack Exchange, Chinese technical communities, official documentation, and general developer web sources for software errors, compatibility problems, API usage questions, and real-world workarounds. Use for debugging and discovery only; results are not paper-citation evidence.","https:\u002F\u002Fgithub.com\u002FCZX-LBJ\u002Fdeepseek-science\u002Fblob\u002Fmain\u002Fpackages\u002Fskill\u002Fskill-science\u002Fassets\u002Fsources\u002Faris\u002Fskills\u002Fweb-debug-search\u002FSKILL.md",[],[]]