BAGEL Script Builder skill
The bagel-script-builder skill from the BAGEL repository turns a protein design goal into a script and a launcher for sweeps.
- 148 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Scripts start heavy computations and cluster submissions that cost money and time.
- AI-generated code needs review before running.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/bagel-script-builderDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Scripts start heavy computations and cluster submissions that cost money and time.
- AI-generated code needs review before running.
This is third-party code. Review the repository files before installing.
What it does
BAGEL, a Python framework for programmable protein design, ships with the `bagel-script-builder` skill. The agent clarifies the design goal from a checklist of questions and writes a verbose, commented script. Every script opens with a note that the code was produced with AI assistance and needs review. For sweeps the skill provides launchers: serial runs, background runs, submission to a SLURM or PBS cluster and parallel runs on Modal. Each configuration goes into its own folder. The repository also holds a second skill, `sae-feature-annotations`.
Who it is for. Protein researchers and bioinformaticians who describe a task in words and want a working BAGEL script instead of assembling energy terms and oracles by hand.
Good fit when
- A binder peptide must be designed for a target across several seeds.
- A set of configurations goes to a SLURM cluster through one launcher command.
- An agent builds a script with ESMFold or Boltz2 oracles from a goal description.
- A team wants to smoke test scripts for one step on Modal before a long run.
Not a fit when
- You need a finished model or a result without running computations.
- There is no access to a GPU cluster or Modal.
- The work is not about proteins, so BAGEL does not fit.
Example request
Use BAGEL to design a 30-residue binder to CD20 and run 8 seeds serially in the background.Limitations
The skill writes scripts, but the computing happens in BAGEL and plugged-in models such as ESMFold, Boltz2 and AlphaFold2, which need installation and strong hardware. A person should review generated code before spending machine time. The README asks users to cite the published BAGEL paper. The skill lives in the repository `.claude/skills/` folder and works with the project after cloning or copying into a personal skills directory.
How to disable. Delete `~/.claude/skills/bagel-script-builder` if you copied the skill. In the repository it sits in `.claude/skills/` and can simply go unused.
FAQ
How do I make the skill global?
Copy the folder with `cp -r .claude/skills/bagel-script-builder ~/.claude/skills/`. Inside the repository Claude Code finds it without copying.
Does it work with other agents?
Yes. `SKILL.md` is self-contained, so ask any agent to read it with the `references/` folder and follow the described flow.
What does the skill produce?
A verbose script with a single `main()`, an optional one-step smoke test on Modal and, for sweeps, a `sweep_runner.py` or `submit_cluster.py` launcher.
Where does BAGEL itself install from?
The recommended route is PyPI, where the package is called `biobagel`. Python 3.12 is required.
What is the second skill?
`sae-feature-annotations` sits beside it in `.claude/skills/` with a `sae_features.py` script and an API reference. Its use is described in that skill `SKILL.md`.
Discussion
No questions or reviews yet.
Related
SkillsNVIDIA BioNeMo Agent Toolkit: skills for life sciences
NVIDIA BioNeMo Agent Toolkit
NVIDIA skills for biology and chemistry: the agent predicts protein structures, docks molecules and designs proteins
A Claude Code agent for protein design: target research, antibody and nanobody generation, screening and candidate ranking.
17 skills for computational biology: RNA-seq, single-cell, protein structure, docking, literature and figures
Probabl skills for machine learning experiments: a scikit-learn and skrub pipeline, evaluation, audit and a link to Skore Hub
A skill that turns a one-line brief into a live 3D factory model with a discrete event simulation, throughput and bottleneck.
24 skills for numerical methods and materials science: the agent checks stability, meshes, convergence and cluster runs.
README in short
The agent section of the README describes the path from a plain-language goal to a script and a sweep. It runs through an interview, API references and launch templates.