jaechang-hits

SciAgent-Skills

# SciAgent-Skills <p align="center"> <img src="https://img.shields.io/badge/skills-199-blue?style=for-the-badge" alt="199 Skills"> <img src="https://img.shields.io/badge/BixBench-92.0%25-brightgreen?style=for-the-badge" alt="BixBench 92.0%"> <img src="https://img.shields.io/badge/license-CC--BY--4.0-lightgrey?style=for-the-badge" alt="CC-BY-4.0"> <img src="https://img.shields.io/github/stars/jaechang-hits/SciAgent-Skills?style=for-the-badge" alt="GitHub Stars"> </p> > **Turn your AI coding agent into a life sciences expert** — 199 bioinformatics skills for Claude Code covering RNA-seq, single-cell analysis, genomics, proteomics, drug discovery, and more. Boosted [BixBench](https://github.com/Future-House/BixBench) from 65% to 92%. Open source. **SciAgent-Skills** is the largest open-source skill library for scientific AI agents. It equips [Claude Code](https://docs.anthropic.com/en/docs/claude-code) (and any markdown-compatible agent) with domain-specific knowledge for computational biology, bioinformatics, cheminformatics, and biostatistics — no fine-tuning required, just plug in and analyze. **Keywords:** bioinformatics AI agent, Claude Code skills, scientific computing, RNA-seq analysis, single-cell RNA-seq, drug discovery pipeline, protein structure prediction, computational biology tools, life science automation, BixBench benchmark ## Benchmark: 92.0% on BixBench-Verified-50 <p align="center"> <img src="assets/benchmark.png" alt="BixBench bioinformatics benchmark results — SciAgent-Skills achieves 92.0% accuracy" width="700"> </p> [BixBench](https://github.com/Future-House/BixBench) is a benchmark for evaluating AI agents on real-world bioinformatics tasks. **SciAgent-Skills achieved 92.0% accuracy** on BixBench-Verified-50, the highest among all tested systems: | System | BixBench-Verified-50 Accuracy | |--------|:----------------------------:| | Claude Code (Opus 4.6) **+ SciAgent-Skills** | **92.0%** | | Claude Code (Opus 4.6) baseline | 65.3% | Simply equipping Claude Code with these domain-specific skills yields a **+26.7 percentage point improvement** — no fine-tuning, no custom model, just structured scientific knowledge. ## Try It Now — OmicsHorizon Want to try these skills without any setup? **[OmicsHorizon](https://omicshorizon.ai/en/)** (오믹스 호라이즌) is the web platform powered by SciAgent-Skills. Sign up and start analyzing your bioinformatics data directly in your browser — RNA-seq, proteomics, drug screening, and more. [](https://omicshorizon.ai/en/) --- **199 ready-to-use scientific skills for AI coding agents** — covering genomics, proteomics, drug discovery, biostatistics, scientific computing, and scientific writing. Each skill is a self-contained SKILL.md file with runnable code examples, key parameters, troubleshooting guides, and best practices. Designed for [Claude Code](https://docs.anthropic.com/en/docs/claude-code), but compatible with any AI agent that reads markdown skill files ([setup guides below](#using-with-other-agents)). ## What's Inside | Category | Skills | Examples | |----------|:------:|----------| | Genomics & Bioinformatics | 65 | Scanpy, BioPython, pysam, gget, KEGG, PubMed, scvi-tools, Bakta, Roary | | Structural Biology & Drug Discovery | 26 | RDKit, AutoDock Vina, ChEMBL, PDB, DeepChem, datamol | | Scientific Computing | 24 | Polars, Dask, NetworkX, SymPy, UMAP, PyG, Zarr, SimPy | | Cell Biology | 15 | pydicom, histolab, FlowIO | | Biostatistics | 12 | scikit-learn, statsmodels, PyMC, SHAP, survival analysis | | Scientific Writing | 21 | Manuscript writing, peer review, LaTeX posters, slides, figure guides | | Systems Biology & Multi-omics | 11 | COBRApy, LaminDB, Reactome, STRING | | Proteomics & Protein Engineering | 10 | ESM, UniProt, PyOpenMS, matchms, HMDB | | Lab Automation | 6 | Opentrons, Benchling | | Data Visualization | 5 | Plotly, Seaborn | | Molecular Biology | 3 | CRISPR sgRNA design, gene expression, cloning | **Skill types:** 72 toolkits, 53 database connectors, 37 guides, 37 pipelines ## Installation ### Prerequisites - An AI coding agent: [Claude Code](https://docs.anthropic.com/en/docs/claude-code), [OpenAI Codex CLI](https://github.com/openai/codex), [Cursor](https://www.cursor.com/), or [Windsurf](https://windsurf.com/) - Git - Python 3.12+ (only needed if you want to run validation scripts) > **Note:** SciAgent-Skills is **not** an npm package. Skills are plain markdown files read directly by your AI agent — no `npx`, `npm install`, or runtime dependencies needed. Just clone the repository and point your agent at the skill files. ### Step 1: Clone the Repository ```bash git clone https://github.com/jaechang-hits/SciAgent-Skills.git cd SciAgent-Skills ``` ### Step 2: Choose Your Setup Method #### Method A: Claude Code Plugin (Recommended) Load SciAgent-Skills as a Claude Code plugin for the current session: ```bash claude --plugin-dir /path/to/SciAgent-Skills ``` To verify the plugin loaded, run `/plugin` inside Claude Code and check that `sciagent-skills` appears in the Installed tab. Skills become available as `/sciagent-skills:<skill-name>`: ``` /sciagent-skills:scanpy-scrna-seq /sciagent-skills:rdkit-cheminformatics /sciagent-skills:pymc-bayesian-modeling ``` Or just describe your task — the agent finds the relevant skill automatically: > "Perform differential expression analysis on this RNA-seq count matrix" **Persistent installation** — to load the plugin automatically in every session, use the plugin install command inside Claude Code: ``` /plugin marketplace add jaechang-hits/SciAgent-Skills /plugin install sciagent-skills ``` #### Method B: Project-Level Integration (Claude Code) Clone into your project directory so Claude Code picks up skills via `CLAUDE.md`: ```bash cd your-project git clone https://github.com/jaechang-hits/SciAgent-Skills.git .sciagent-skills ``` Add to your project's `CLAUDE.md`: ```markdown ## Scientific Skills Reference skills in `.sciagent-skills/skills/` for domain-specific analysis. Registry: `.sciagent-skills/registry.yaml` ``` #### Using with Other Agents SciAgent-Skills works with any AI agent that can read project files. Clone the repo into your project, then configure the agent to discover skills via `registry.yaml`. **Method C: OpenAI Codex CLI** ```bash cd your-project git clone https://github.com/jaechang-hits/SciAgent-Skills.git .sciagent-skills cp .sciagent-skills/integration-templates/AGENTS.md ./AGENTS.md ``` Codex reads `AGENTS.md` at the project root automatically. If you already have an `AGENTS.md`, append the contents from the template. **Method D: Cursor** ```bash cd your-project git clone https://github.com/jaechang-hits/SciAgent-Skills.git .sciagent-skills mkdir -p .cursor/rules cp .sciagent-skills/integration-templates/cursor-rules.md .cursor/rules/sciagent-skills.md ``` Cursor loads rules from `.cursor/rules/` automatically. Alternatively, you can use the `AGENTS.md` template — Cursor supports it as well. **Method E: Windsurf** ```bash cd your-project git clone https://github.com/jaechang-hits/SciAgent-Skills.git .sciagent-skills mkdir -p .windsurf/rules cp .sciagent-skills/integration-templates/windsurf-rules.md .windsurf/rules/sciagent-skills.md ``` Windsurf loads rules from `.windsurf/rules/` automatically. Alternatively, you can use the `AGENTS.md` template — Windsurf supports it as well. **Other agents**: If your agent reads project files, clone the repo as `.sciagent-skills/` and instruct the agent (via its config mechanism) to read `.sciagent-skills/registry.yaml` for skill discovery. ### Step 3: Install Dependencies ```bash cd SciAgent-Skills pixi install ``` [Pixi](https://pixi.sh) handles the Python environment and all required packages. If you don't have pixi installed: ```bash curl -fsSL https://pixi.sh/install.sh | bash ``` ## How Skills Work Each skill follows a structured template: ``` skills/<category>/<skill-name>/ SKILL.md # Main skill file (300-550 lines) references/ # Optional deep-dive reference files assets/ # Optional templates, configs ``` A **SKILL.md** contains: - **Frontmatter** — name, description, license (for agent discovery) - **Overview & When to Use** — what the tool does and when to reach for it - **Prerequisites** — packages, data, environment setup - **Quick Start** — minimal copy-paste example - **Workflow / Core API** — step-by-step pipeline or module-by-module API guide - **Key Parameters** — tunable settings with defaults and ranges - **Common Recipes** — self-contained snippets for common tasks - **Troubleshooting** — problem/cause/solution table The agent reads only the `description` field during planning. Full skill content is loaded on demand when relevant. ## Directory Structure ``` SciAgent-Skills/ ├── .claude-plugin/ │ └── plugin.json # Claude Code plugin manifest ├── integration-templates/ # Config templates for Codex, Cursor, Windsurf ├── skills/ # All 199 skills organized by category │ ├── genomics-bioinformatics/ │ ├── structural-biology-drug-discovery/ │ ├── scientific-computing/ │ ├── cell-biology/ │ ├── biostatistics/ │ ├── scientific-writing/ │ ├── systems-biology-multiomics/ │ ├── proteomics-protein-engineering/ │ ├── lab-automation/ │ ├── data-visualization/ │ └── molecular-biology/ ├── templates/ # Skill authoring templates ├── registry.yaml # Index of all skills ├── CLAUDE.md # Skill authoring guide └── scripts/ └── validate_registry.py ``` ## Example Use Cases **Single-Cell RNA-seq Analysis** (scRNA-seq) > "Load 10X data, QC filter, normalize, cluster, find marker genes, and annotate cell types" Uses: `anndata-data-structure` → `scanpy-scrna-seq` **Bulk RNA-seq Differential Expression** > "Run DESeq2-style differential expression analysis on this count matrix, generate volcano plot" Uses: `pydeseq2-differential-expression` → `matplotlib-scientific-plotting` **Drug Discovery Pipeline** > "Search ChEMBL for EGFR inhibitors with IC50 < 100nM, filter with Lipinski rules using RDKit, dock top candidates with AutoDock Vina" Uses: `chembl-database-bioactivity` → `rdkit-cheminformatics` → `autodock-vina-docking` **Protein Structure Prediction & Analysis** > "Get the AlphaFold structure for UniProt P04637, assess confidence, find high-confidence binding regions" Uses: `uniprot-protein-database` → `alphafold-database-access` **Bayesian Biostatistics** > "Fit a hierarchical Bayesian model to this clinical trial data with patient-level random effects" Uses: `pymc-bayesian-modeling` → `matplotlib-scientific-plotting` **Multi-omics Integration** > "Integrate transcriptomics and proteomics data, run pathway enrichment, build a protein interaction network" Uses: `lamindb-data-management` → `reactome-pathway-analysis` → `string-protein-interaction` ## Contributing ### Adding a New Skill 1. Read `CLAUDE.md` for the full authoring workflow 2. Classify your topic (pipeline / toolkit / database / guide) 3. Pick a category from the table above 4. Use the appropriate template from `templates/` 5. Add the entry to `registry.yaml` 6. Validate: `python scripts/validate_registry.py` ### Skill Templates | Template | Use When | |----------|----------| | `SKILL_TEMPLATE.md` | Linear input→process→output pipeline (e.g., DESeq2) | | `SKILL_TEMPLATE_TOOLKIT.md` | Collection of independent modules (e.g., RDKit) | | `SKILL_TEMPLATE_PROSE.md` | Conceptual guide, decision frameworks (e.g., statistical test selection) | ## Requirements - Python 3.12+ (for validation scripts) - No runtime dependencies — skills are markdown files read by the agent - Individual skills list their own tool-specific prerequisites (e.g., `pip install scanpy`) ## Comparison with Other Tools | Feature | SciAgent-Skills | GPTomics/bioSkills | ClawBio | |---------|:-:|:-:|:-:| | Total skills | **199** | ~30 | ~20 | | BixBench benchmark | **92.0%** | — | — | | Skill types | Pipeline, Toolkit, Database, Guide | Pipeline | Pipeline | | Multi-agent support | Claude Code, Codex, Cursor, Windsurf | Claude Code | Claude Code | | Claude Code plugin | Yes | No | No | | Web platform | [OmicsHorizon](https://omicshorizon.ai/en/) | No | No | ## License CC-BY-4.0 for original content. Individual skills note the license of their underlying tools. ## Acknowledgments This project builds on 50+ open-source scientific Python packages. If you find a skill useful, consider starring the underlying tool's repository and supporting its maintainers. --- _{**Related searches:** bioinformatics AI agent, Claude Code scientific skills, RNA-seq analysis tool, single-cell RNA-seq AI, drug discovery AI pipeline, protein structure prediction, computational biology automation, life science AI tools, 바이오인포매틱스 AI, 오믹스 호라이즌, 생명과학 AI 에이전트, BixBench benchmark}