New: 98,000+ skills indexed from GitHub repos with 100+ stars · every skill safety-checkedBrowse top skillsTop skills
Esc to closeBrowse all skills
Submit a skillBrowse skills
Submit a skillBrowse skills

Agent skill

Biopipelines

by locbp-uzh in locbp-uzh/biopipelines

Design and run computational protein and ligand workflows on a GPU: binder and enzyme design, de novo backbone generation, inverse folding and sequence redesign, structure prediction, protein-ligand…

MITAuto-check passedResearch & Science

Install Biopipelines

skills CLI
$ npx skills add locbp-uzh/biopipelines --skill biopipelines -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install locbp-uzh/biopipelines biopipelines --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/locbp-uzh/biopipelines.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/biopipelines .claude/skills/biopipelines && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
biopipelines
GitHub stars
109
Token cost
~2.4k tokens
SKILL.md length
1,033 words
Files
9 (incl. references)
Skills in repo
1
Repo updated
First seen
Licence
MIT

At a glance

Design and run computational protein and ligand workflows on a GPU: binder and enzyme design, de novo backbone generation, inverse folding and sequence redesign, structure prediction, protein-ligand…

  • Wants to design
  • SKILL.md covers First: is bp-mcp registered?, Finding tools without the MCP…, The framework contract and The API in one screen, plus 2 more sections
  • Runs Python scripts from its folder; calls python, ssh and pip
  • Analyze proteins

What it does

Biopipelines is an agent skill from locbp-uzh/biopipelines. Design and run computational protein and ligand workflows on a GPU: binder and enzyme design, de novo backbone generation, inverse folding and sequence redesign, structure prediction, protein-ligand docking, compound-library and covalent screening, saturation mutagenesis, stability and solubility engineering, codon optimization. BioPipelines (locbp-uzh, CSBJ 2026) puts 86 tools behind one declarative Pipeline API - RFdiffusion 1/2/3, BoltzGen, ProteinMPNN, LigandMPNN, LASErMPNN, Frame2Seq, ThermoMPNN, AlphaFold…

Its SKILL.md is about 2.4k tokens, which your agent loads only when the skill is triggered. The skill folder holds 9 other files, including reference files (for example `build_tool_index.py`, `draft_tool_tags.py` and `references/cluster_backend.md`).

It sits in Research & Science, covering Protein structure and design and Drug discovery and cheminformatics. It works with RDKit and AlphaFold. The repository describes itself as: A modular and user-friendly Python framework for protein and ligand engineering workflows on SLURM clusters. The licence is MIT.

When your agent uses it

  • Wants to design
  • Analyze proteins
  • Names one of these tools
  • Has a BioPipelines pipeline to write

Example prompts

  • “/biopipelines”

Requirements

  • Python 3
  • Docker

What it can do on your machine

Read from SKILL.md and the folder at commit 675e466. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Ships script files (Python), which the agent can run.

    Shell commands in SKILL.md call:

    • python
    • ssh
    • pip

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    Links to these hosts (documentation or services it may open):

    • spj.science.org

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Biopipelines loads about 2.4k tokens when it runs, and up to ~24k if it reads all its reference files. Until then it costs about 235 tokens; SKILL.md has 1,033 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~235
When it runs · the whole SKILL.md, loaded when a task matches
~2.4k
With references · SKILL.md plus every file in references/, read only if the agent opens them
~24k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from locbp-uzh/biopipelines at commit 675e466, republished under its MIT licence (© locbp-uzh). 1,033 words, ~2,395 tokens.

Download SKILL.mdSave it as .claude/skills/biopipelines/SKILL.md (or your agent's skills folder). This skill also uses 8 other files; get the full folder from GitHub.
name
biopipelines
description
Design and run computational protein and ligand workflows on a GPU: binder and enzyme design, de novo backbone generation, inverse folding and sequence redesign, structure prediction, protein-ligand docking, compound-library and covalent screening, saturation mutagenesis, stability and solubility engineering, codon optimization. BioPipelines (locbp-uzh, CSBJ 2026) puts 86 tools behind one declarative Pipeline API - RFdiffusion 1/2/3, BoltzGen, ProteinMPNN, LigandMPNN, LASErMPNN, Frame2Seq, ThermoMPNN, AlphaFold, Boltz2, ESMFold2, DiffDock, GNINA, Vina, NeuralPLexer, OpenMM, FPocket, P2Rank, PLIP, ProLIF, PoseBusters, Prodigy, RDKit, ADMET-AI - and tracks IDs through typed data streams so a multi-stage campaign stays traceable end to end. Load when the user wants to design, predict, dock, screen, redesign or analyze proteins or ligands, names one of these tools, or has a BioPipelines pipeline to write, run or debug.

BioPipelines

BioPipelines (Quargnali & Rivera-Fuentes, LOC-BP UZH; CSBJ 10.34133/csbj.0129) is a Python framework that puts 86 protein/ligand-modeling tools behind one declarative Pipeline API. You describe a workflow as a chain of tools; the framework generates and runs the per-tool scripts, tracks IDs through typed data streams, and materializes outputs.

This skill ships inside the repo, so adding it to any agent is one step and it tracks the framework as it evolves. It covers using the framework, and running it on any single-node GPU backend (Modal/RunPod, a Docker GPU box, or an interactive Slurm GPU shell) via the generic container config variant.

First: is bp-mcp registered?

Look at your own tool list. If it holds bp_tools, bp_setup, bp_runs, bp_status, bp_logs, bp_lineage, bp_table, bp_submit, bp_resubmit, bp_cancel, bp_visualize, bp_fetch, bp_project, bp_provenance and bp_reproduce, those tools are the interface to BioPipelines and the file-reading and shell recipes in this skill are what a session without them falls back on.

What you wantCallInstead of
Find a toolbp_tools(), bp_tools(tags=[...]), bp_tools(outputs="rmsd")reading references/tool_index.md
A tool's parametersbp_tools(name="LigandMPNN")opening docs/tool/*.md
Run a pipelinebp_submit(script=...)ssh s3it "./submit ..."
Status, logs, tables, lineagebp_status, bp_logs, bp_table, bp_lineagessh + ls + tail + scp
Resume, stop, or show a runbp_resubmit, bp_cancel, bp_visualize./resubmit, scancel, bp-visualize over ssh
Project documentsbp_project(project_dir=..., action=...)writing PROJECT.md by hand

Do not run a script in this repo, or a shell command, to do what one of these does. skills/biopipelines/build_tool_index.py in particular is a maintainer script that regenerates the catalog file; it is not how you look a tool up. If no bp_* tool is present, follow the sections below, and see references/mcp_server.md to set the server up.

Finding tools without the MCP server — read the index, not the prose docs

references/tool_index.md is your catalog. 88 callable entries, one line each: name, version, hardware (CPU/GPU), the platforms it is verified on, what it does, and a pointer to its full entry. It costs ~4k tokens; the prose docs it points into cost ~78k, and the README's badge table another ~32k.

The loop is: read the index → pick your tools → read only those sections. Each entry ends with docs/tool/<file>.md#<anchor> — read that section for the real signature before you write the stage.

Each index line ends with the tool's tags — a closed 38-term vocabulary in four facets, defined in docs/tool_tags.md. Grep them: covalent, binder-design and motif-scaffolding are the sort of capability the one-line summary never mentions.

Two rules, because both failure modes are silent:

  • Never read a whole docs/tool/*.md file to find a tool. analysis.md alone is 24k tokens for 33 tools, and a campaign typically uses three of them.
  • Never guess a tool name or a parameter. An unknown keyword is swallowed by **kwargs rather than rejected, so a typo becomes a silently ignored argument, not an error.

The index lists 88 entries against 86 registered tools: SolubleMPNN and LoadMultiple are public classes that share a parent's identity (ProteinMPNN and Load), so they are callable but carry no separate registration.

The framework contract

llm/ is the authoritative, author-maintained agent contract — it overrides anything paraphrased here. Read the file that matches what you are doing; do not read them all up front.

  • llm/pipelines.md — authoring a Pipeline: the tool/data-stream model, ID tracking, combinatorics. Read this before writing a multi-stage workflow.
  • llm/development.md — only when changing framework code, not when using it.
  • references/cluster_backend.md, references/colab_backend.md, references/daint_backend.md, references/container_backend.md (in this skill) — one backend reference each. Read the one you are on. They live here rather than in llm/ so any host can use them, not only a session that was told to read llm/.
  • docs/user_manual.md — the long-form manual. Consult sections as needed; it is ~17k tokens end to end.
  • references/mcp_server.md (in this skill) — installing and registering bp-mcp, which serves bp_tools.
Show full SKILL.md (414 more words)Show less

The API in one screen

python
from biopipelines import Pipeline, Resources, Sequence, Ligand, UniProt, Boltz2

with Pipeline("Project", "job_name", description="..."):
    Resources(gpu="A100", memory="64GB", time="6:00:00", cpus=8)
    prot = Sequence("MSEQ...", type="protein")           # or UniProt("Q15436") to fetch it
    lig  = Ligand(smiles="C[N+]1=C(...)...")
    Boltz2(proteins=prot, ligands=lig, output_format="mmcif")

Every tool takes its inputs by keyword. Boltz2's first positional parameter is config (a raw YAML string), so Boltz2(prot, lig, ...) binds the protein to config and the ligand to proteins — always write proteins= / ligands=.

Each Resources(...) call opens a new batch (one scheduler job), inheriting whatever it does not specify from the previous batch. Call it once per resource profile — a CPU-only stage after a GPU stage gets its own Resources(gpu="none", ...) rather than inheriting the GPU and walltime of the heaviest stage. Inside a plain Parallel() block every sibling iteration must call Resources() to open its own batch; inside Parallel(pack=N) it is called once to describe the whole node allocation and Run(...) delimits each task.

Tool.install() is called inside the with Pipeline(...) block (a bare Boltz2.install() at module scope is a silent no-op) — or just use bp-warm, which wraps it for you.

Running on any single-node GPU host (the container backend)

The repo ships a generic single-node variant so you do not hand-author a config per provider. Set three env vars and point one config line at your persistent mount:

bash
export BIOPIPELINES_CONFIG_VARIANT=container   # select config.container.yaml
export BIOPIPELINES_OTF=1                       # run tools inline (no scheduler)
export BIOPIPELINES_LOCAL_OUTPUT=0              # honor configured output dir, NOT cwd

In config.container.yaml edit folders.base.root: to your persistent mount (default /workspace); the config's own paths — home, data, scratch, weight caches, biopipelines_output — all derive from it. The one path that does not is the micromamba env root: that is MAMBA_ROOT_PREFIX, hardcoded to /workspace/micromamba in Dockerfile.container, so a different root: needs MAMBA_ROOT_PREFIX=<root>/micromamba exported alongside it (see references/container_backend.md). First, warm the tools you need onto that mount once:

bash
bp-warm Boltz2 ProteinMPNN        # builds per-tool micromamba envs + downloads weights
python my_pipeline.py             # subsequent runs reuse the warm env + cached weights

Dockerfile.container in the repo root builds a ready image (CUDA 12.4 + micromamba + pip install -e '.[colab]', with build-essential/gcc present — several tools JIT-compile a CUDA/C helper at import and fail without it).

Why BIOPIPELINES_LOCAL_OUTPUT=0 matters: with OTF on a non-Colab scheduler the framework otherwise diverts output to the ephemeral ./outputs (cwd), silently overriding your configured biopipelines_output. On a container that directory is lost at teardown. Setting it to 0 routes results to the persistent mount.

Reporting back

Check the completion markers before reporting anything as successful: they are empty files named <NNN>_<ToolName>_COMPLETED / _FAILED / _WARNING, written one level above each tool's output folder, and the per-tool log is <Job>/Logs/<NNN>_<ToolName>.log.

Save the structure (.cif/.pdb), the confidence/affinity JSON, and a summary figure as artifacts. For a co-fold, report pTM, ipTM/ligand-ipTM, complex pLDDT, and affinity_probability_binary (affinity defaults to True). Note that Boltz2 has no covalent-mechanism knowledge: an unconstrained co-fold of a covalent ligand finds a non-covalent pocket, not the reactive residue — use Boltz2's covalent_linkage constraint when the mechanism is covalent.

© locbp-uzh, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 8 other files (references) in skills/biopipelines of locbp-uzh/biopipelines.

  • SKILL.md
  • build_tool_index.py
  • draft_tool_tags.py
  • references/cluster_backend.md
  • references/colab_backend.md
  • references/container_backend.md
  • references/daint_backend.md
  • references/mcp_server.md
  • references/tool_index.md

Open the folder on GitHubat commit 675e466

Compare with similar skills

Biopipelines next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Biopipelines compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Biopipelines this skilllocbp-uzh/biopipelines109—~2.4kAutomated safety check: PassMIT
Esm Protein Language Modeljaechang-hits/SciAgent-Skills3741 repos~4kAutomated safety check: PassMIT
DiffDock Molecular DockingK-Dense-AI/scientific-agent-skills48k1 repos~3kAutomated safety check: NotesMIT
Chai1JimLiu/science-skills2284 repos~1.2kAutomated safety check: PassApache-2.0
Emdb Databasejaechang-hits/SciAgent-Skills3741 repos~4.9kAutomated safety check: PassCC-BY-4.0
Pdb Databasejaechang-hits/SciAgent-Skills3741 repos~7.7kAutomated safety check: PassBSD-3-Clause

Similar skills

  • Esm Protein Language Model

    jaechang-hits/SciAgent-Skills

    Protein language models (ESM3, ESM C) for sequence generation, structure prediction, inverse folding, and embeddings.

    374 GitHub starsUsed in 1 repo~4k tokens
    Research & ScienceAuto-check passed

  • DiffDock Molecular Docking

    K-Dense-AI/scientific-agent-skills

    Predicts how small molecules bind to a protein with DiffDock, covering batch docking, pose ranking by confidence and checks on the results; not for binding affinity.

    48k GitHub starsUsed in 1 repo~3k tokens
    Research & ScienceAuto-check: notes

  • Chai1

    JimLiu/science-skills

    Structure prediction for protein, nucleic-acid, and small-molecule complexes with the Chai-1 foundation model (Chai Discovery 2024, github.com/chaidiscovery/chai-lab).

    228 GitHub starsUsed in 4 repos~1.2k tokens
    Research & ScienceAuto-check passed

  • Emdb Database

    jaechang-hits/SciAgent-Skills

    Look up EMDB cryo-EM density maps and fitted atomic models via the entry REST API + EBI Search WS.

    374 GitHub starsUsed in 1 repo~4.9k tokens
    Research & ScienceAuto-check passed

  • Pdb Database

    jaechang-hits/SciAgent-Skills

    Query RCSB PDB (200K+ structures) via the public REST + GraphQL APIs with plain requests (no SDK).

    374 GitHub starsUsed in 1 repo~7.7k tokens
    Research & ScienceAuto-check passed

  • Alphafold Database Fetch And Analyze

    google-deepmind/science-skills

    Retrieve and analyze AlphaFold predicted structures for a protein.

    3.2k GitHub starsUsed in 2 repos~1.2k tokens
    Research & ScienceAuto-check passed

Works with

Categories

Questions about Biopipelines

What does Biopipelines do?

Design and run computational protein and ligand workflows on a GPU: binder and enzyme design, de novo backbone generation, inverse folding and sequence redesign, structure prediction, protein-ligand…. Biopipelines is an agent skill from locbp-uzh/biopipelines. Design and run computational protein and ligand workflows on a GPU: binder and enzyme design, de novo backbone generation, inverse folding and sequence redesign, structure prediction, protein-ligand docking, compound-library and covalent screening, saturation mutagenesis, stability and solubility engineering, codon optimization.

When should I use Biopipelines?

Biopipelines fits situations like: wants to design; analyze proteins; names one of these tools; has a BioPipelines pipeline to write.

How do I install Biopipelines in Claude Code?

Run `npx skills add locbp-uzh/biopipelines --skill biopipelines -a claude-code`. Or copy the skill folder (skills/biopipelines in locbp-uzh/biopipelines) into .claude/skills/biopipelines in your project. Claude Code loads it when a task matches its description.

How do I install Biopipelines in Codex?

Run `npx skills add locbp-uzh/biopipelines --skill biopipelines -a codex`. Or copy the skill folder (skills/biopipelines in locbp-uzh/biopipelines) into .agents/skills/biopipelines in your project. Codex loads it when a task matches its description.

Can I use Biopipelines in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add locbp-uzh/biopipelines --skill biopipelines -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/biopipelines, .gemini/skills/biopipelines, .github/skills/biopipelines and .opencode/skills/biopipelines in your project.

What does Biopipelines need to run?

Going by SKILL.md and its folder, Biopipelines needs Python for the scripts in its folder and the command-line tools its instructions call (python, ssh and pip). Our summary lists: Python 3; Docker.

Does Biopipelines access the network?

SKILL.md names 1 domain. As links in the text: spj.science.org. This is read from the text; nothing was executed.

Is Biopipelines safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Biopipelines use?

Biopipelines is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Biopipelines use?

About 2.4k tokens (SKILL.md is roughly 9.6k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 22k tokens, read only when the agent opens those files.

What are the alternatives to Biopipelines?

Skills that share tags, products or a category with Biopipelines: Esm Protein Language Model (jaechang-hits/SciAgent-Skills, 374 stars), DiffDock Molecular Docking (K-Dense-AI/scientific-agent-skills, 48k stars), Chai1 (JimLiu/science-skills, 228 stars) and Emdb Database (jaechang-hits/SciAgent-Skills, 374 stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Biopipelines?

locbp-uzh (a GitHub organization) maintains it in locbp-uzh/biopipelines, which has 109 GitHub stars. The repository was last updated on September 30, 2026.

Source: locbp-uzh/biopipelines on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.