1.3K skills · sorted by relevance
Track ML experiments with automatic logging, visualize training in real-time, optimize hyperparameters with sweeps, and manage model registry with W&B - collaborative MLOps platform
The English summary is being prepared. Showing the original description.
Use a live Jupyter kernel for stateful, iterative Python execution via hamelnb. Load this skill when the task involves exploration, iteration, or inspecting intermediate results — data science, ML experimentation, API exploration, or building up complex code step-by-step. Uses terminal to run CLI commands against a live Jupyter kernel. No new tools required.
The English summary is being prepared. Showing the original description.
Design precise, production-grade enterprise UI — dashboards, admin panels, SaaS interfaces, and data-heavy applications. Use this skill for professional software interfaces that need Jony Ive-level craft: clean, minimal, functional, with intentional personality. Do NOT use for creative landing pages, marketing sites, or experimental/artistic frontends.
The English summary is being prepared. Showing the original description.
Transforma análise de desempenho em plano de ação priorizado por tiers — ganhos rápidos, mudanças estruturais, experimentos e o que parar de fazer. Use quando o usuário disser "o que eu faço agora", "como melhorar", "plano de ação", "por onde começo", "otimizar campanha", ou logo após perf-analyzer e perf-patterns. Fecha o ciclo: escrever → medir → aprender → escrever melhor.
The English summary is being prepared. Showing the original description.
Static-to-EDS conversion that preserves the original design while making content authorable in Document Authoring. Two modes — page-level (overlay template with slot markers) and block-level (each section becomes an independent EDS block). Use when converting an AI-generated static HTML page (Stardust, Mobirise, Relume, Lovable, v0, Figma-derived hand-coded, etc.) into an Edge Delivery Services page. Triggers on "convert this page to EDS", "static-to-EDS overlay", "convert to EDS blocks", "next experimentation", "next run", "start run", or when a user provides a source URL and asks to make it editable in DA while keeping the original design intact. Do NOT use for canonical EDS block-rewrite migrations — that's the page-import skill.
The English summary is being prepared. Showing the original description.
The original v1 taste-skill, preserved for projects depending on its exact behavior. The current default is `design-taste-frontend` (v2 experimental), which is a substantial rewrite. Use this v1 install name only if you need exact backward compatibility.
The English summary is being prepared. Showing the original description.
Sanguo RAG unresolved mention resolution loop. Use for: 三國 RAG, unresolved mentions, alias review, observed mentions, MCQ triage, manual roster seeds, false-positive exclusion, running run_resolution_loop.py, applying unresolved-triage answers, delegating 三國名詞查證 to a web-capable agent, or Pilot ETL chunk experiments with LangChain text splitters / LangGraph / LangFlow around 張飛長坂橋等小樣本流程。
The English summary is being prepared. Showing the original description.
Design a self-improvement loop for a shipped product's AI agents — production traces feed evals, evals feed improvement proposals (prompts, playbooks, retrieval configs), and a human approval gate promotes changes with rollback. Technique-agnostic: chooses per project between ACE-style evolving playbooks, GEPA/MIPROv2 offline optimization, or simple eval-driven iteration via references/techniques.md. Load when the user asks to make my product's agents self-improving, learn from production traces, add a learning loop, evolve prompts or playbooks safely, promote agent improvements, or GEPA-style optimization. NOT harness-evolution (that improves the coding agent), NOT experimentation (product A/B tests), NOT agent-run-retro (dev-phase manual retros — this skill is the production-scale continuation).
The English summary is being prepared. Showing the original description.
Archive a completed change in the experimental workflow. Use when the user wants to finalize and archive a change after implementation is complete.
The English summary is being prepared. Showing the original description.
Azure Weights & Biases SDK for .NET. ML experiment tracking and model management via Azure Marketplace. Use for creating W&B instances, managing SSO, marketplace integration, and ML observability.
The English summary is being prepared. Showing the original description.
Compare cell-type composition and spatial features across conditions in IMC/MIBI cohorts with the patient as the experimental unit, covering pseudoreplication, per-patient aggregation, mixed models, compositional (Dirichlet/scCODA) differential abundance, diffcyt, per-image-to-patient spatial differential testing (SpaceANOVA), batch covariates, and FDR. Use when testing whether a cell type or spatial niche differs between groups, avoiding cell-level pseudoreplication, choosing a differential-abundance method, or correctly powering an IMC cohort comparison.
The English summary is being prepared. Showing the original description.
Differential abundance (DA) and differential state (DS) analysis for flow and mass cytometry - tests which cell populations change in frequency or marker expression between conditions using diffcyt (edgeR/voom/GLMM for DA, limma/LMM for DS), with cydar, CITRUS, and compositional methods (sccomp, scCODA, DCATS) as alternatives. Covers the sample-is-the-experimental-unit principle, design/contrast and mixed-model formulas, compositionality of cluster proportions, and FDR across clusters. Use when comparing populations between groups, choosing a DA method, handling paired/batch designs, or deciding whether compositional correction is needed.
The English summary is being prepared. Showing the original description.
Controls error rates across thousands of simultaneous tests in genomics discovery using false-discovery-rate methods (Benjamini-Hochberg 1995; Benjamini-Yekutieli 2001 for arbitrary dependence; Storey q-value with pi0 estimation; local FDR; independent filtering Bourgon 2010; covariate-weighted FDR via IHW Ignatiadis 2016), plus family-wise error control (Bonferroni, Holm) and the GWAS genome-wide threshold. Covers the FDR-versus-FWER choice as the discovery-versus-confirmatory distinction, the dependence assumptions behind BH (PRDS) versus BY, pi0 estimation, the independent-filtering and false-coverage-rate traps, and reproducibility ranking via IDR (Li 2011). Use when correcting p-values from genome-wide tests, choosing between BH/BY/q-value/Bonferroni, setting an FDR threshold, applying IHW or independent filtering, or interpreting q-values. For confirmatory trials with few pre-specified endpoints (closed testing, graphical/gatekeeping), see clinical-biostatistics/multiplicity-graphical.
The English summary is being prepared. Showing the original description.
Build weighted gene co-expression networks to identify modules of co-regulated genes, relate them to phenotypes, and find hub genes using WGCNA, hdWGCNA, MEGENA, CEMiTool, and Gaussian graphical models. Covers signed-network choice, soft-threshold selection, module preservation, and the marginal-vs-partial-correlation distinction. Use when finding co-expression modules, identifying hub genes, relating gene networks to clinical or experimental traits, or building single-cell co-expression networks. For directed TF-target inference see scenic-regulons and grn-inference; for condition rewiring see differential-networks.
The English summary is being prepared. Showing the original description.
Cross-method decision tree for calling hits in pooled CRISPR screens. Catalogs statistical models (MAGeCK RRA, MAGeCK MLE, BAGEL2, drugZ, JACKS, Chronos, CERES), experimental designs each is built for, failure modes outside design domain, reconciliation when methods disagree, multiple-testing and effect-size thresholds, the order of operations (count -> QC -> CN-correct -> hit-call -> validate), the second-best-sgRNA conservative rule, and consensus-hit strategy. Use when choosing among MAGeCK / BAGEL2 / drugZ / JACKS / Chronos for a given design, reconciling disagreement across two or three methods on the same screen, deciding whether to require consensus, gating downstream validation by hit-confidence tier, or interpreting unstable hit lists across reruns.
The English summary is being prepared. Showing the original description.
Designs and analyzes in vivo CRISPR screens in animal tumor models, organoids, and immune-cell adoptive transfers. Covers bottleneck math (250x cells/sgRNA requires ~25M cells implanted; impossible for most syngeneic models, forcing focused libraries), focused library design (Manguso 2017 Nature 547:413 immune screen; Chen 2015 tumor screens), CRISPR-StAR intrinsic-control screening (Uijttewaal 2025 Nat Biotechnol 43:1848), clonal-dynamics-limited detection, tumor-explant DNA recovery, syngeneic vs xenograft vs PDX considerations, and the relationship to downstream MAGeCK / drugZ analysis. Use when designing in vivo CRISPR screens for tumor / immune / metastasis biology, choosing focused vs genome-wide for animal models, addressing bottleneck-induced clonal collapse, picking the syngeneic / xenograft / PDX model, integrating in vivo with in vitro results, or applying CRISPR-StAR for animal experiments.
The English summary is being prepared. Showing the original description.
Analyzes pooled CRISPR screens with MAGeCK (Li et al 2014), covering count generation (mageck count), the RRA two-condition workflow (mageck test using alpha-RRA over per-sgRNA negative-binomial p-values), the MLE multi-condition workflow (mageck mle with explicit design matrix and beta-score output), normalization choice (median vs total vs control-sgRNA vs spike-in), sgRNA efficiency injection, paired-sample testing, time-course design, drug-screen versus dropout-screen design matrices, MAGeCKFlute and MAGeCK-VISPR downstream visualization, and decision logic for when to use MAGeCK vs JACKS / BAGEL2 / drugZ / Chronos. Use when running a fresh CRISPR screen analysis, picking RRA vs MLE for the experimental design, choosing a normalization method from QC signatures, debugging MLE convergence failure or NaN beta scores, comparing MAGeCK output across tools, or building a batch-aware multi-cell-line / multi-condition MLE design matrix.
The English summary is being prepared. Showing the original description.
Analyzes single-cell pooled CRISPR screens (Perturb-seq, CROP-seq, Perturb-CITE-seq, ECCITE-seq, multiome) where each cell carries an sgRNA and a scRNA-seq / surface-protein / chromatin readout. Covers experimental design (direct-capture Perturb-seq Dixit 2016 vs CROP-seq 3'UTR-barcoded Datlinger 2017 vs ECCITE-seq vs Multiome), MOI for sgRNA assignment, escaper-cell filtering (Mixscape, Papalexi 2021), SCEPTRE NB GLM + permutation for low-MOI (Barry 2024 Genome Biol 25:124), the Pertpy framework, factor decomposition, genome-scale Perturb-seq (Replogle 2022 Cell, 2.5M cells), and per-perturbation single-cell DE. Use when running a single-cell CRISPR screen, choosing direct-capture vs CROP-seq architecture, filtering escaper cells, performing single-cell DE, integrating Perturb-seq with pathway analysis, scaling to GW CRISPRi via Replogle protocol, or analyzing multi-omics screens.
The English summary is being prepared. Showing the original description.
Identify differentially bound regions across CLIP-seq conditions (knockdown vs control, treatment vs vehicle, disease vs healthy) using DEWSeq (sliding-window DESeq2), Flipper (Skipper-downstream), ASpeak, edgeR, or limma-voom. Use when computing condition-level changes in RBP binding intensity, choosing peak-level vs window-level vs crosslink-level testing, designing replicate experiments, or distinguishing biological binding shifts from technical confounders.
The English summary is being prepared. Showing the original description.
Predict RBP binding from RNA sequence using deep learning models (RBPNet sequence-to-signal, RNAProt RNN, GraphProt2 GCN with structure, DeepCLIP, DeepRiPe multi-modal CNN) for variant-effect prediction, in silico binding-site discovery, model interpretation, and transfer learning from CLIP and RBNS datasets. Use when computational prediction of RBP binding from sequence is needed, evaluating variant effects on binding without further wet-lab experiments, comparing model performance, or training a custom model on ENCODE eCLIP data.
The English summary is being prepared. Showing the original description.
Visualizes ChIP-seq data using deepTools (computeMatrix, plotHeatmap, plotProfile, bamCoverage, bamCompare), pyGenomeTracks (modern INI-driven track plots), Gviz (R browser-style), EnrichedHeatmap (ComplexHeatmap-based), ChIPseeker tag heatmaps, and IGV batch screenshots. Handles bigWig normalization choices (CPM, BPM, RPGC, spike-in scaled), bamCompare operations (log2 ratio, subtract, SES), k-means clustering of heatmaps for biological subgrouping, and spike-in-scaled tracks for global-shift experiments. Use when generating publication-quality ChIP-seq signal heatmaps, profile plots, genome-browser tracks, or comparing samples visually.
The English summary is being prepared. Showing the original description.
Identifies differentially bound ChIP-seq regions between conditions using DiffBind, csaw (sliding windows), DESeq2/edgeR/PyDESeq2 on count matrices, NormR (control-aware), or MAnorm2. Distinguishes three distinct normalization problems (composition bias, trended bias, global shifts) and matches each to its appropriate fix including spike-in scaling. Use when comparing ChIP-seq binding between experimental conditions, choosing normalization for global vs local changes, integrating spike-in data, or reconciling DiffBind/DESeq2 disagreement.
The English summary is being prepared. Showing the original description.
Designs PROTACs, molecular glues, and bivalent degraders with explicit handling of E3 ligase choice (VHL, CRBN, IAP, MDM2, KEAP1), linker design (length, composition, rigidity), ternary complex prediction (PRosettaC, DeepTernary, AlphaFold3 with constraints), cooperativity (alpha), DC50 / Dmax characterization, hook effect, and prediction-experiment reconciliation. Use when designing targeted protein degraders, planning linker SAR, predicting ternary complex stability, or building generative degrader workflows.
The English summary is being prepared. Showing the original description.
Use when introducing a feature flag system, choosing between release flags / experiment flags / operational flags / kill switches / permission flags, designing percentage rollouts, building kill-switch + circuit-breaker patterns, structuring evaluation contexts (user IDs, plan tier, region), or building flag-cleanup discipline. Triggers: LaunchDarkly / GrowthBook / Unleash / Flagsmith / OpenFeature, percentage ramp, sticky bucketing, kill switch wired to alerting, "we still have flags from 2 years ago", flag debt, OpenFeature provider, evaluation context, default value on provider failure. NOT for A/B test statistical analysis (separate skill), traffic-split routing at the edge, build-time bundler defines, or environment-based config.
The English summary is being prepared. Showing the original description.