メニュー
Use when an Evo 2 phage SFT project needs a reproducible Microviridae or new phage genome collection from NCBI, paper-linked repositories, supplements, or other public biological databases.
デフォルトでは、最初にソースを確認する Prompt が選択されています。直接コマンドに切り替えるか、ローカルコピーをダウンロードすることもできます。
インストールを決める前に、SKILL.md と SkillsMP に表示されている付属ファイルをお読みください。
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/NVIDIA-BioNeMo/bionemo-recipes --skill bionemo-phage-design-collect-genomesコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
Use when a phage-design workflow must discover or adapt to local GPU, SSH, Slurm, Lepton, manual, or unfamiliar execution infrastructure and produce durable launch, monitoring, resume, or handoff commands.
Use after selecting an Evo 2 phage SFT checkpoint and defining RL objectives to calibrate prompt serialization, temperature, prefix-length distribution, and fixed validation sampling.
Use when producing, deduplicating, hard-QC screening, clustering, ranking, or selecting final phage designs from a chosen RL checkpoint.
SKILL.md を表示中
| name | bionemo-phage-design-collect-genomes |
| description | Use when an Evo 2 phage SFT project needs a reproducible Microviridae or new phage genome collection from NCBI, paper-linked repositories, supplements, or other public biological databases. |
| metadata | {"author":"NVIDIA <bionemofeedback@nvidia.com>"} |
Work inside the recipe and result roots selected by the controller. Build a licensed, documented collection of usable full genomes.
Define the taxon and host scope, completeness/topology requirements, length and ambiguity limits, record-status policy, exclusions, and license constraints before collection. For the Microviridae replication, inspect and use the recipe's maintained downloader. For a new target, use the research skill and follow primary sources or database records to the relevant sequence files.
Validate and document the collection using the concise collection guidance. Normalize stable record IDs without losing their source IDs, and keep conflicting host annotations separate rather than guessing.
Do not stop collection merely because an arbitrary count was reached. Assess whether the corpus has enough distinct full genomes and similarity clusters, nucleotide/token mass, target-relevant coverage, length/composition support, and cluster-held-out validation/test support for the chosen model and training strategy. Expand the search, use transfer learning, or revise the model plan when the evidence is inadequate. Never inflate support with exact duplicates, circular rotations, accession versions, or interaction rows.
Write the unprefixed FASTA collection, metadata table, exclusions, and source notes under the stage artifacts, plus a concise SUMMARY.md and RUNLOG.md. When a target is selected, record its FASTA path, accession, and topology for SFT preparation.
