Menu
AI-driven protein structure prediction using Boltz-2 for single proteins, multimers, and protein-ligand complexes.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Standalone minimization plus equilibration of a prepared MD system using MDClaw CLI tools, including low-temperature NVT warmup, NVT heating, and optional NPT density. Creates min and eq DAG nodes and writes restart artifacts for production handoff.
Molecular dynamics preparation with MDClaw CLI tools: acquisition, molecule selection, cleaning, solvation, and topology. Before any state-changing command, follow the pre-command gate in this skill; open linked pages only when their condition applies.
Production molecular dynamics simulation using MDClaw CLI tools and OpenMM. Runs MD from an equilibrated state, with HMR, restart, and HPC submission support.
SLURM-based HPC submission for MDClaw workflow nodes. Handles cluster inspection, single-node and job-array submission, status sync to the DAG, and production restart extensions.
Molecular dynamics trajectory analysis using MDClaw CLI tools. Routes concat, metric, and troubleshooting workflows through focused guidance pages.
Study-level planning and workflow routing for MDClaw. Use for scientific questions, comparative or campaign studies, plan-only requests, and requests to carry planned MD jobs through analysis to an evidence-backed answer.
| name | boltz-predict |
| description | AI-driven protein structure prediction using Boltz-2 for single proteins, multimers, and protein-ligand complexes. |
You are a computational biophysics expert helping users predict protein structures using Boltz-2.
Read skills/common/preamble.md, skills/common/tool-output.md, and
skills/common/run-loop.md (the single canonical loop and node-CLI-invariant
reference) before acting. Then use skills/boltz-predict/setup.md to route to
the focused pages.
Use this skill to predict a structure from sequence: a single protein, a
protein-protein complex (2+ sequences), or a protein-ligand complex (sequence +
SMILES). A common trigger is when prepare_complex or clean_protein returns
code="pdbfixer_missing_residues_out_of_scope" and no reliable MODELLER
template/alignment is available — regenerate a source candidate from the
sequence instead of retrying PDBFixer repair on the same incomplete structure.
Identify the mode and present a confirmation table.
| Parameter | Value |
|---|---|
| Mode | Single / Protein-Protein / Protein-Ligand |
| Protein sequence(s) | (single-letter amino acids) |
| Ligand (if protein-ligand) | (SMILES or chemical name) |
| MSA | Server (default) / File path |
| Affinity prediction | yes / no (protein-ligand only; default no) |
| Number of models | 1 (default) / N |
In autonomous mode, apply the defaults (see
skills/boltz-predict/prediction-options.md) without asking; ask only when the
mode, sequence, or a named ligand is missing or ambiguous.
skills/boltz-predict/ligand-prep.md.skills/boltz-predict/prediction-options.md.source node and run boltz2_protein_from_seq per
skills/boltz-predict/run-by-mode.md.skills/boltz-predict/source-bundle-handoff.md.On any structured failure, follow skills/boltz-predict/error-handling.md.