Build and maintain an LLM-curated personal knowledge base — the "LLM Wiki" pattern from Andrej Karpathy's April 2026 gist. Use this skill whenever the user wants to ingest a source (paper, article, transcript, PDF, notes) into a persistent compounding knowledge base, ask a question against accumulated notes, lint or audit such a base, or initialize a new one. Trigger on phrases like "add this to my wiki", "ingest this paper", "compile this into the knowledge base", "what does my wiki say about X", "lint the wiki", "build a knowledge base from these documents", "research notes", "second brain", "personal knowledge base", or any reference to LLM Wiki / OmegaWiki. Trigger even when the user does not say "wiki" — if they are accumulating sources over time and want them organized, this applies. The skill scales — sharded indexes, atomic pages, YAML frontmatter, and a bundled search script keep the wiki from becoming a context bottleneck at hundreds or thousands of pages.
Ingest ONE newly-added or changed `docs/` file into the project LLM Wiki at `wiki/`, routing by ASPECT — each doc group is ingested differently. Use this whenever a single source doc is added/changed and its knowledge must land in the wiki, whatever its aspect: an algorithm spec (`docs/algorithms/**`), a validation artifact (`docs/Evidence/**`), a validation verdict (`docs/Validation/**`), a testing methodology (`docs/checklists/**`), an MCP-layer doc (`docs/mcp/**`), or a top-level/refactoring/skills/templates doc. Adding one algorithm typically drops SEVERAL such files (spec + Evidence + report + MCP tool) — run this per file; each is routed by its own group rules, not bundled blindly. Trigger phrases: "ingest this doc into the wiki", "I added <file>, update the wiki", "process the new Evidence/report/spec/tool", "reflect this validation report in the wiki", "wire the new MCP tool into the wiki". Do NOT use this for the multi-hundred-file backfill — that is the `/wiki:ingest` loop over `tools/wiki-ingest/ne
Predict and analyze RNA secondary structure with Seqeron (MCP tools OR the C# API). Use to fold an RNA / compute minimum free energy (MFE, Turner 2004 nearest-neighbour, kcal/mol), predict the secondary structure (dot-bracket + base pairs), enumerate stem-loops / hairpins, detect pseudoknots (crossing base pairs), find antiparallel complementary arms (RNA inverted repeats), classify base pairs (Watson-Crick / G-U wobble) and test if two bases can pair, parse / validate dot-bracket (WUSS) notation, and compute individual loop / stem / dangling-end / terminal-mismatch free-energy terms. Triggers: "fold this RNA", "predict secondary structure", "minimum free energy of…", "MFE of this RNA", "find stem-loops / hairpins", "detect pseudoknots", "is this dot-bracket valid", "parse this dot-bracket", "can G and U pair", "what base-pair type", "hairpin/internal/bulge/multibranch loop energy". Server: analysis (all RnaSecondaryStructure.*).
Build phylogenetic trees and compute population-genetics statistics with Seqeron (MCP tools OR the C# API). Phylogenetics: distance matrices (p-distance / Jukes-Cantor / Kimura-2P / Hamming), tree construction (Neighbor-Joining, UPGMA), Newick parse/serialize, tree stats (depth, length, leaves, MRCA, patristic + Robinson-Foulds distance), bootstrap support. Population genetics: allele/genotype frequencies, minor allele frequency (MAF) + MAF filtering, Hardy-Weinberg equilibrium test, Fst / F-statistics / pairwise Fst, nucleotide diversity (π), Watterson's θ, Tajima's D, heterozygosity, linkage disequilibrium (D'/r²) + haplotype blocks, runs of homozygosity / inbreeding, iHS, ancestry, selection scans. Triggers: "build a tree", "make a phylogeny / dendrogram from these sequences", "neighbor-joining / UPGMA tree", "distance matrix", "compute Fst", "test Hardy-Weinberg / HWE", "nucleotide diversity of this population", "Tajima's D", "linkage disequilibrium", "MAF filter". Servers: phylogenetics + population.
Design and QC molecular-biology reagents with Seqeron's MolTools server (47 tools) — PCR primer pairs and hybridization probes, CRISPR guide RNAs with off-target and specificity scoring, codon optimization of a CDS for a target organism (CAI/GC before-after), restriction-site finding and digest simulation, and melting-temperature (Tm) with salt correction. Use for "design primers for this template", "find/design CRISPR guides for this target and check off-targets", "codon-optimize this gene for E. coli / yeast / human", "where does EcoRI cut / simulate a digest", "design a qPCR/FISH probe", "compute the salt-corrected Tm", primer-dimer / hairpin / 3'-stability QC, and choosing compatible restriction enzymes. Dual-mode: MCP tool calls + equivalent Seqeron.Genomics C# Method IDs.
Annotate and characterize genomic sequences with Seqeron (MCP tools OR the C# API). Use to find ORFs / predict genes / promoters / RBS, call + annotate + classify variants (SNPs, indels, VEP-like effect, ACMG-like pathogenicity), discover and scan motifs (exact/degenerate/PROSITE/ PWM), analyze repeats & low-complexity / masking, profile k-mers & composition, plus splicing. (Read-evidence structural variants / CNVs → seqeron-structural-variants; miRNA targets / seeds / pre-miRNA hairpins → seqeron-mirna; RNA secondary-structure folding → seqeron-rna-structure; protein-feature prediction → seqeron-protein-features; methylation/epigenetics → seqeron-epigenetics; transcriptome/RNA-seq → seqeron-transcriptome; ANI/orthologs/synteny/comparative genomics → seqeron-comparative-genomics.) Triggers: "annotate this sequence", "find ORFs / genes / promoters", "call variants", "what motifs are in…", "classify this variant", "predict the effect", "mask low-complexity", "find repeats", "k-mer profile". Servers: annotation
Assemble sequencing reads into contigs and QC the result with Seqeron (MCP tools OR the C# API). Use to assemble reads via de Bruijn graph or overlap-layout-consensus (OLC), build/merge/scaffold contigs, compute read coverage over a reference, and report assembly statistics / N50; plus k-mer counting, k-mer spectrum, most-frequent / min-count / unique k-mers and clumps for assembly QC, and assembly-relevant repeat & low-complexity screening (direct/inverted/tandem repeats, palindromes, microsatellites, longest repeat). Triggers: "assemble these reads", "de Bruijn / OLC assembly", "compute N50 / assembly stats", "coverage of these reads over the reference", "k-mer spectrum of…", "most frequent k-mers", "find clumps", "screen for repeats before assembly". Servers: analysis + core (k-mer/repeat subset); the assembly-graph engine (de Bruijn/OLC/coverage/N50) is on the Alignment server.
One-time install & configuration for a freshly cloned Seqeron repo, so the bioinformatics skills are ready to solve real tasks. Use when the user says "install and configure", "set up Seqeron", "get me started", "onboard me", "prepare the repo", "why aren't the bio tools working", or asks how to make the Seqeron MCP tools / skills available. Verifies the .NET 10 SDK and Python are present, builds all 11 MCP servers into the on-demand cache (so the first real tool call is instant), and runs a live smoke test — without registering MCP anywhere (tool schemas stay out of the model's context; skills pull tools in on demand). Run this once per clone/machine; after it, just describe a biology task.