Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add javidmardanov/paper-banana-skill --skill paper-bananagit clone --depth 1 https://github.com/javidmardanov/paper-banana-skillWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/javidmardanov/paper-banana-skill/paper-banana)<a href="https://agentmods.dev/skills/javidmardanov/paper-banana-skill/paper-banana"><img src="https://agentmods.dev/badge/skills/javidmardanov/paper-banana-skill/paper-banana/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/javidmardanov/paper-banana-skill/paper-banana"><img src="https://agentmods.dev/badge/skills/javidmardanov/paper-banana-skill/paper-banana.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00146 | $0.04061 |
| Opus 5.5 | $0.00058 | $0.01624 |
| Sonnet 5.5 | $0.00029 | $0.00812 |
| Haiku 4.5 | $0.00015 | $0.00406 |
Grade A, and why
paper-banana scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 301 lines — stays where its author put it; the contents beside it link to each section on GitHub.
PaperBanana: Academic Illustration Pipeline
Automates publication-ready academic illustrations via 5 specialized agents, each a separate call to the configured VLM or image model (Gemini by default): Retriever (categorize & select references) -> Planner (multimodal description) -> Stylist (polish) -> Visualizer (render) -> Critic (evaluate & refine).
Two output modes:
- DIAGRAM MODE: Each agent is a Python script calling the configured VLM / image model. Run
scripts/orchestrate.pyfor end-to-end execution. - PLOT MODE: Statistical plots generated as executable Python matplotlib/seaborn code (code-based to eliminate data hallucination).
Requirements: Python 3.10+ with google-genai>=2, matplotlib, seaborn, numpy, pillow, and an API key for the provider in use (default Gemini: GOOGLE_API_KEY or GEMINI_API_KEY). Optional: openai (for openai/... and openrouter/... models), anthropic, pypdf (PDF input).
Models: named provider/model; unprefixed names are inferred. Defaults are gemini-3.5-flash (Retriever/Planner/Stylist/Critic) and gemini-3-pro-image (Visualizer, "Nano Banana Pro"). Override with PAPERBANANA_VLM_MODEL / PAPERBANANA_IMAGE_MODEL or the orchestrator's --vlm-model / --image-model flags.
| Provider | Key | Reasoning | Image | Examples |
|---|---|---|---|---|
| gemini (default) | GOOGLE_API_KEY |
yes | yes | gemini-3.5-flash, gemini-3.1-flash-lite, gemini-3-pro-image, gemini-3.1-flash-image |
| openai | OPENAI_API_KEY |
yes | yes | openai/gpt-5.5, openai/gpt-5.4-mini, openai/gpt-image-2, openai/gpt-image-1.5 |
| anthropic | ANTHROPIC_API_KEY |
yes | no | anthropic/claude-opus-5, anthropic/claude-sonnet-5 |
| openrouter | OPENROUTER_API_KEY |
yes | yes | openrouter/google/gemini-3.5-flash, openrouter/google/gemini-3-pro-image |
If the user names a provider or model, pass it via --vlm-model / --image-model. If a run fails with a missing-key or model error, run python scripts/validate_output.py --check-api and report which key or model is the problem.
Paper: PaperBanana: Automating Academic Illustration for AI Scientists (arXiv:2601.23265, Google/PKU). Official code: https://github.com/dwzhu-pku/PaperBanana
What ships with it
36 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- assets/matplotlib_styles/academic_default.mplstyle 1.4 KB
- assets/matplotlib_styles/academic_minimal.mplstyle 1.4 KB
- assets/palettes/colorblind_safe.json 1.0 KB
- assets/palettes/neurips_methodology.json 1.5 KB
- assets/palettes/neurips_plots.json 1.0 KB
- assets/references/2404.15806v1.jpg 107 KB
- assets/references/2601.03570v1.jpg 28 KB
- assets/references/2601.05110v1.jpg 75 KB
- assets/references/2601.05144v1.jpg 19 KB
- assets/references/2601.06411v1.jpg 51 KB
- assets/references/2601.06953v2.jpg 60 KB
- assets/references/2601.07033v1.jpg 39 KB
- assets/references/2601.07055v1.jpg 58 KB
- assets/references/2601.09259v1.jpg 94 KB
- assets/references/2601.09708v1.jpg 86 KB
- assets/references/2601.14724v2.jpg 98 KB
- assets/references/2601.15165v2.jpg 30 KB
- assets/references/2601.15892v2.jpg 51 KB
- assets/references/index.json 5.8 KB
- assets/venues.json 1.7 KB
- references/DIAGRAM-CATEGORIES.md 3.3 KB
- references/DIAGRAM-PROMPTS.md 10 KB
- references/DIAGRAM-STYLE-GUIDE.md 7.4 KB
- references/EVALUATION-RUBRIC.md 5.8 KB
- references/PLOT-PROMPTS.md 8.3 KB
- references/PLOT-STYLE-GUIDE.md 6.8 KB
- scripts/common.py 5.0 KB runs code
- scripts/critic.py 6.7 KB runs code
- scripts/generate_image.py 7.3 KB runs code
- scripts/orchestrate.py 18 KB runs code
- scripts/planner.py 7.3 KB runs code
- scripts/plot_generator.py 15 KB runs code
- scripts/providers.py 16 KB runs code
- scripts/retriever.py 6.4 KB runs code
- scripts/stylist.py 6.6 KB runs code
- scripts/validate_output.py 12 KB runs code
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago Changed · +55 lines · +37 tokens per session f273730f4c87
- 1mo ago First seen · 246 lines · 109 tokens per session scan A 86f63acd366a
paper-banana is a skill published in the GitHub repository javidmardanov/paper-banana-skill (3 stars, last pushed 1mo ago), licensed MIT. It adds 146 tokens to every session and 4,061 once invoked, about $0.0006 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
social-physics-filter
Unified boundary enforcement, interpersonal diagnostic, and relational audit engine. Absorbs 40 psychology + 2 social protocols and all relationship case studies.
arxiv-mcp-server
Use when finding, comparing, reading, or monitoring arXiv papers, including requests for abstracts, citation graphs, original LaTeX, section-level technical details, or literature reviews.
distribution-physics
Analyzes market dynamics and go-to-market strategies using "Distribution First" architecture.
polish-sci-figures
A final review and preparation workflow for scientific figures, including charts, image panels, manuscripts, slides, and posters.
make-sci-data-figures
A workbench for turning raw CSV, TSV, or Excel tables into defensible scientific figures. It chooses charts based on the biological question and data structure rather than assuming one chart type fits every dataset.
standardize-sci-images
Scientific image standardization for microscopy, fluorescence, histology, pathology, and electron-microscopy image batches. Use when the user has raw or derived scientific raster images needing non-destructive processing, locked tone settings, equal dimensions, calibrated scale bars, manifests, montages…