技能创建器
skill-creator
创建、改进和评估自定义技能,把专业流程转化为可重复使用的 AI 能力。
data-and-outputs【数据与研究】Validate OmegaFold FASTA inputs, inspect pseudo-MSA tensors, predict PDB output names, and interpret confidence stored as B-factors.
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
Use this sub-skill when an agent needs to validate OmegaFold FASTA input, explain how pipeline.fasta2inputs converts sequences into pseudo-MSA cycles, predict output PDB file names, or interpret confidence values stored in PDB B-factor columns.
For end-to-end model inference commands, use inference CLI. For model construction, configs, and OmegaFold.forward, use model API.
scripts/inspect_fasta_pipeline.py; it imports omegafold.pipeline, runs CPU-only fasta2inputs, and never downloads weights or runs the model.references/data-formats.md.references/api-reference.md.references/troubleshooting.md.Run the bundled helper before full inference when input/output behavior is unclear:
python sub-skills/data-and-outputs/scripts/inspect_fasta_pipeline.py --write-tiny-pdb
The helper creates a tiny FASTA unless --fasta is provided, prints sequence count, normalized residue indices, p_msa/p_msa_mask shapes for every cycle, predicted PDB paths, and optionally writes a tiny synthetic PDB through pipeline.save_pdb.
This sub-skill owns FASTA parsing, pseudo-MSA input preparation, output file naming, save_pdb, atom14 writer expectations, and confidence-as-B-factor explanations. It intentionally does not cover weight downloads, accelerator selection, resource flags, or full prediction commands; route those to inference CLI.