What problem does it solve?
Designing new protein binders against a target requires authoring complex API payloads, choosing scaffold and crop configurations, sizing expensive generation campaigns, and ranking thousands of outputs. This Skill guides the full Boltz protein:design workflow so binders are generated, costed, and ranked correctly.
Core Features & Use Cases
- Payload authoring and submission: Builds protein:design payloads with target normalization, binder_specification variants (boltz_curated, structure_template, no_template), sequence DSL, and optional design rules, then runs estimate-cost and start through the boltz-api CLI.
- Target exploration: Scouts crop radii, disorder cutouts, domains, and binding sites with 50-design runs using bundled gemmi/numpy scripts before committing to a full campaign.
- Cost gating and ranking: Enforces an explicit spending gate before every paid run and ranks downloaded results by binding_confidence with iptm and min_interaction_pae tiebreakers.
- Use Case: A researcher wants nanobody binders against a new multi-domain target. The Skill recommends an exploration pass, scouts domain and site configurations in parallel, picks the winner by max binding_confidence, then submits a 50,000-design campaign after cost confirmation.
Quick Start
Use the boltz-protein-design skill to design nanobody binders against my target CIF file, starting with a target-exploration pass and confirming the estimated cost before submitting.
