Jason-0409-G

vivarium-compare

Compare a set of microbial genomes: genome relatedness (ANI/AAI), shared vs unique gene content (orthology / pangenome / core-accessory via OrthoFinder), and genome structure (synteny / rearrangements via MUMmer). Use whenever the user wants to know how similar two or more genomes are, whether two strains are the same species, which genes are shared or unique to a strain, build orthogroups or a pangenome, or align genomes to see synteny. Triggers on phrases like "compare these genomes", "ANI/AAI between", "are these the same species", "which genes are unique to strain X", "core and accessory genome", "run OrthoFinder", "pangenome", "synteny / genome alignment", "比较这些基因组", "算 ANI/AAI", "是不是同一个种", "哪些基因是 X 独有的", "核心/附属基因组", "共线性". Light steps (FastANI, MUMmer) run locally in the bio_tools conda env and produce tables ready for vivarium-report; the heavy step (OrthoFinder) is scaffolded as a ready-to-run command. Part of the vivarium comparative-genomics skill set.

vivarium-phylo

Build a phylogeny from sequences and test genes for selection. Use whenever the user wants to align sequences, build a gene or species tree, infer phylogeny, bootstrap a tree, or test for positive/purifying selection (dN/dS, ω) with PAML. Triggers on phrases like "build a tree", "phylogeny / phylogenetic tree of", "align these sequences", "MAFFT / trimAl / IQ-TREE", "bootstrap support", "is this gene under selection", "dN/dS / omega", "run PAML / codeml", "建树/系统发育树", "比对这些序列", "建一棵树", "自展支持", "这个基因受不受选择", "算 dN/dS", "跑 PAML". Alignment→trim→tree runs locally in the bio_tools conda env; large trees and PAML selection tests are scaffolded as ready-to-run commands. Part of the vivarium comparative-genomics skill set.

vivarium-prep

Get genomes ready for comparative analysis: assess assembly quality (contigs, N50, GC, length, completeness), and annotate genes and function (Prokka, eggNOG, dbCAN/CAZy). Use whenever the user wants genome statistics or QC, to check how good an assembly is, to assemble long reads, to annotate a genome, to call genes, or to get COG/KEGG/CAZy function tables before comparing genomes. Triggers on phrases like "genome stats / QC", "what's the N50 / GC / contig count", "how good is this assembly", "annotate this genome", "run Prokka / eggNOG / dbCAN", "call genes", "assemble these reads", "基因组质控/统计", "N50/GC/contig 数", "组装质量怎么样", "注释这个基因组", "跑 Prokka/eggNOG/dbCAN", "CAZy/COG/KEGG 注释". Light QC runs locally in the bio_tools conda env; heavy steps (assembly, eggNOG, dbCAN) are scaffolded as ready-to-run commands. Part of the vivarium comparative-genomics skill set.

vivarium-report

Make publication-grade figures (and tables / methods text) from comparative-genomics results, in Python OR R. Use whenever the user wants to plot or visualize genomics analysis output — an ANI/AAI heatmap, a CAZy/COG/KEGG category bar chart, a pangenome or presence/absence matrix, a dN/dS (ω) plot, a phylogenetic tree, a synteny/dot plot — or to turn analysis tables into a journal-ready figure, or draft a methods paragraph with tool versions. Triggers on phrases like "plot the ANI matrix", "make a heatmap of", "bar chart of CAZy/COG categories", "figure for the paper", "make this publication quality", "出个图", "画个热图/柱状图", "把这个表画成图", "出版级/Nature 风格的图", "ANI 热图". Produces editable SVG + PDF + TIFF (600 dpi) with Nature-style conventions (Arial, small fonts, editable text, restrained palette). This skill plots results that already exist — to *compute* ANI/AAI use vivarium-compare, to *build* a tree use vivarium-phylo, to *run* a search use vivarium-search; this one turns their output into the figure. Part of the v

vivarium-search

Local sequence-similarity search for comparative genomics. Use whenever the user wants to find homologs of a gene or protein, BLAST or DIAMOND a query against a genome/proteome/CDS, check whether a genome or proteome carries a homolog of some sequence, identify what an unknown sequence is, find the best hit for a protein, or do an HMM/profile domain search (Pfam/dbCAN/TIGRFAM). Triggers on phrases like "blast this", "find homologs of", "search this protein against", "does this genome have a copy of", "what is this sequence", "best hit for", "搜一下这个序列/基因", "比对/BLAST 一下", "这个基因组有没有 X 的同源", "找同源", "鉴定这条序列", and any local protein/nucleotide similarity or domain search. Runs locally in the bio_tools conda env (blastp/blastn/tblastn/blastx, diamond, hmmer); returns a tidy named-column hit table, an interpretation, and a reproducible command. Part of the vivarium comparative-genomics skill set.

vivarium

Orchestrate an end-to-end comparative-genomics workflow over a genome set. Use for multi-stage or complete analyses such as "run the whole pipeline", "characterize these strains", "compare these genomes and make figures", "跑完整比较基因组流程", or requests naming vivarium 2.0, durable execution, the event ledger, crash recovery, or the full goal. Default new end-to-end work to the 2.0 durable plan/run engine; use the 1.0 mutable-manifest orchestrator only when the user explicitly requests legacy V1 behavior. For one isolated ANI, BLAST, tree, or figure, defer to the corresponding vivarium sub-skill.