ESMFold2

SkillProductivity

Lets your agent predict quick protein structures from amino acid sequences with confidence scores.

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the ESMFold2 skill

About this capability

Predict quick protein structures with ESMFold-style workflows. Use when a task needs fast MSA-free folding, sequence triage, variant structure screening, or confidence review.

What this skill tells your AI

The instructions your AI receives, as published by companion-inc/feynman in skills/esmfold2/SKILL.md and read by ahel’s review.

Use this skill when a fast protein fold hypothesis is useful before heavier structure prediction.

Workflow:

  1. Normalize sequences into FASTA and record identifiers, mutations, truncations, domains, and oligomer assumptions.
  2. Verify the local or endpoint route before running.
  3. Save FASTA, model version, command or request body, predicted PDB/mmCIF, pLDDT/confidence outputs, and logs.
  4. Flag low-confidence regions, missing multimers, disorder, membrane regions, and sequence lengths outside the chosen route's limits.
  5. Use RDKit/3Dmol/PDB previews and source-backed comparisons when the output influences a research decision.

Use ESMFold-style predictions for triage unless an independent check supports the structural claim.

Signals

GitHub stars
10k
Forks
1k
Last commit
Sep 2026
Advanced
Catalog kind
skill
Gateway key
esmfold2-companion-inc
Source
github.com/companion-inc/feynman