AgentFigureGallery

AgentFigureGallery is a drop-in scientific plotting skill for Claude Code, Codex, Cursor, and other coding agents.
It turns real visual references plus human like/reject feedback into action-ready plotting guidance before code is written.

One-command bootstrap for Codex:

curl -fsSL https://raw.githubusercontent.com/Dsadd4/AgentFigureGallery/main/scripts/install.sh | bash

Clone once, run one command, and your coding agent gets a human-curated scientific figure gallery before writing plotting code. Codex, Claude Code, and Cursor-compatible installs are supported.

agent query -> gallery display -> human like/reject/select -> agent action

AgentFigureGallery helps coding agents stop guessing what a publication figure should look like. The agent queries visual references first, the human marks taste preferences in a browser gallery, and the selected references are exported as an action bundle before plotting code is written.

Codex Skill Smoke Test

After installing the Codex skill, AgentFigureGallery can render one Nature-style smoke example for every supported plot type.

See examples/plot_type_examples/ for the runnable script, source data, and PNG/PDF/SVG outputs.

Install

git clone https://github.com/Dsadd4/AgentFigureGallery.git
cd AgentFigureGallery
python -m venv .venv
source .venv/bin/activate
pip install -e .
agentfiguregallery doctor
agentfiguregallery install-skill --target codex
agentfiguregallery gallery --plot-type embedding_plot --limit 50 --serve

Install all agent entrypoints:

curl -fsSL https://raw.githubusercontent.com/Dsadd4/AgentFigureGallery/main/scripts/install.sh | AFG_AGENT_TARGETS="codex claude-code cursor" bash

For Agents

After pip install -e . finishes, tell your Codex, Claude Code, Cursor, or other coding agent:

Read skills/agent-figure-gallery/SKILL.md, then use AgentFigureGallery before writing publication figure code.

Or install the agent skill wrapper first:

agentfiguregallery install-skill --target codex
agentfiguregallery install-skill --target claude-code
agentfiguregallery install-skill --target cursor
agentfiguregallery install-cursor-rule --project /path/to/your-cursor-project

Codex installs to ~/.codex/skills, Claude Code installs to ~/.claude/skills, Cursor-compatible installs to ~/.cursor/skills, and Cursor Project Rules install to .cursor/rules/agent-figure-gallery.mdc. See docs/AGENT_QUICKSTART.md and examples/agent_prompt.md.

End-to-end examples:

• examples/end_to_end_embedding.md
• examples/generated_embedding_plot/README.md
• examples/before_after_benchmark/README.md

Full public KB:

agentfiguregallery setup --pack full-public --manifest-url https://huggingface.co/datasets/dsadd4/AgentFigureGallery/resolve/main/resource_manifest.json

Fallback when Hugging Face is blocked:

agentfiguregallery setup --pack full-public --manifest manifests/resource_manifest.github-api.json

Dynamic Gallery

Use the browser gallery to generate candidates by plot type, remove bad references globally, keep type-specific preferences, and export selected references for the agent that will write the final plotting code. Every like/reject becomes reusable taste memory, so the system gets sharper as humans and agents keep using it.

agentfiguregallery query --task "Nature-style embedding map for cell atlas"
agentfiguregallery gallery --plot-type embedding_plot --limit 100 --serve

What Is Inside

• 16,341 full-public visual candidates across 10 scientific plot types.
• Glike-curated minimal pack committed for instant smoke tests.
• Codex-equipped plot-type smoke examples with PNG/PDF/SVG outputs.
• Backend CLI, browser gallery, Codex skill wrapper, and agent expansion guide.
• Candidate IDs, global/type-level preferences, and export bundles for agent handoff.

Roadmap

• Curated Cell and Science style reference packs
• Faster full-public mirror for China and restricted networks

Completed proof point:

• One-command Codex skill install
• Generated embedding plot from a reference bundle
• Before/after benchmark: prompt-only vs reference-guided plotting

Docs

• ExtendAgent/: instructions for agents that expand the gallery.
• docs/AGENT_QUICKSTART.md: minimal instructions for coding agents.
• docs/DISCOVERY_PLAYBOOK.md: launch and star-growth checklist.
• docs/releases/v0.1.0.md: first public release notes.
• docs/HF_SYNC.md: Hugging Face dataset card and asset sync commands.
• docs/PYPI_RELEASE.md: Python package release path.
• docs/HF_DATASET_CARD.md: Hugging Face dataset card draft.
• docs/LAUNCH.md: public launch copy and channels.
• docs/FULL_KB_DISTRIBUTION.md: public asset-pack strategy.
• docs/REMOTE_FULL_VALIDATION.md: first remote full-public validation and current mirror-speed caveat.