What problem does it solve?
Long-running Boltz jobs (structure predictions, ADME, screens, designs) can outlive an agent session, leaving users unsure of job state or how to recover already-computed results without resubmitting. This Skill provides a disciplined workflow for listing, retrieving, and resuming downloads of existing Boltz jobs.
Core Features & Use Cases
- Job Discovery: List recent jobs across all six Boltz resources (structure-and-binding, ADME, small-molecule/protein screens and designs), merged and sorted by creation time.
- Status Retrieval: Route
retrieve calls by job ID prefix (e.g., sab_pred_*, prot_des_*) and read per-resource progress counters.
- Result Recovery: Resume interrupted downloads with
download-results using the saved idempotency key as the run name, with cursor-based idempotent resume via .boltz-run.json.
- Use Case: Your session died mid-download of a protein design job. Use this Skill to inspect local checkpoint state with
download-status, recover the idempotency key via retrieve, and resume the download into the original run directory without re-submitting the job.
Quick Start
Use the boltz-check-status skill to find my Boltz job with ID prot_des_abc123, check its status, and resume downloading its results into my experiments folder.
