Machine Learning Skills

Orchestrate Xcode build optimization by benchmarking first, running the specialist analysis skills, prioritizing findings, requesting explicit approval, delegating approved fixes to xcode-build-fixer, and re-benchmarking after changes. Use when a developer wants an end-to-end build optimization workflow, asks to speed up Xcode builds, wants a full build audit, or needs a recommend-first optimization pass covering compilation, project settings, and packages.

Prepares and validates GPT-RAG umbrella releases across component repositories and the AI Landing Zone. Use for manifest pins, changelog entries, release branches, tags, and GitHub Release notes.

Use when designing how one canonical agent core runs across Codex, Claude Code, Gemini CLI, Cursor, and AGENTS.md-compatible tools.

AWS Bedrock foundation models for generative AI. Use when invoking foundation models, building AI applications, creating embeddings, configuring model access, or implementing RAG patterns.

Building applications with Large Language Models - prompt engineering, RAG patterns, and LLM integration. Use for AI-powered features, chatbots, or LLM-based automation.

Iteratively inspect an agent repository and optional user-provided traces, interview the user, and create, run, and audit Harbor evals one at a time. Use for agent evals, Harbor tasks, benchmark cases, verifier design, or controlled agent environments.

INVOKE THIS SKILL when building ANY retrieval-augmented generation (RAG) system. Covers document loaders, RecursiveCharacterTextSplitter, embeddings (OpenAI), and vector stores (Chroma, FAISS, Pinecone).

> Create and validate Earth2Studio data source wrappers (DataSource, ForecastSource, DataFrameSource, ForecastFrameSource) from remote stores. Do NOT use for fetching data with existing sources, model inference, or installation tasks.

> Fetch weather/climate data via Earth2Studio data sources for specific variables and times. Do NOT use for inference pipelines, model discovery, or installation.

> Build deterministic forecast scripts with Earth2Studio (model, data source, IO, inference). Do NOT use for ensemble, diagnostics, data-only fetch, or install.

> Find Earth2Studio models, data sources, and examples for a weather/climate use case. Do NOT use for writing inference code, downloading data, or installation.

> Guide installing Earth2Studio via uv or pip, selecting model extras, and configuring the environment. Do NOT use for writing inference code, choosing models, or PhysicsNeMo questions.

>- Adversarially review a diff, patch, or plan for memtier_benchmark using the real review standards of the project's senior maintainers (Yossi Gottlieb / yossigo, Oran Agra / oranagra, Paulo Sousa / paulorsousa). Use when asked to "adversarially review", "review like the maintainers", "find what a reviewer would block on", or before opening/merging a PR. Emits skeptical, evidence- backed findings; assumes a problem is real until it can be refuted.

Find Japanese NLP GitHub repositories and Hugging Face models/datasets that are NOT yet in awesome-japanese-nlp-resources. Suggests candidates to add for a given topic using WebSearch + WebFetch, then outputs contribution-ready markdown.

Search all Japanese NLP resources (libraries, models, datasets, tutorials, dictionaries, Hugging Face). Accepts keywords or natural language questions in any language.

Given a Japanese NLP GitHub repo or Hugging Face model/dataset (URL / owner/repo / tool name), find repositories or models/datasets that do the same or related processing. Mines the bundled dataset for content-similar items, then expands via web research across both GitHub and Hugging Face, then merges and re-ranks.

Iteratively optimizes a cuDF RapidsUDF implementation for GPU performance. Use after testing and benchmarking with udf-benchmark. Runs a loop of profiling, optimizing, testing, and benchmarking until performance converges or the iteration budget is exhausted.

Assists with benchmarking and profiling the performance of an Apache Spark UDF on the GPU. This is step 3 of 3 in the UDF conversion workflow (udf-gen-test -> udf-convert-to-* -> udf-benchmark). Use this skill when you have a CPU UDF and a RapidsUDF or SQL implementation, and need to benchmark the performance of the CPU UDF against the GPU implementation.

Update the third_party/kineto submodule in PyTorch to the latest commit from this kineto repo and commit the change. Use when updating the kineto submodule hash for a release.

Build private, on-device AI features on iPhone, iPad, and Mac with Foundation Models, Core ML, MLX Swift, or llama.cpp. Use when choosing an Apple-local model runtime, building an Apple Intelligence chatbot or tool-calling feature, running an LLM on Apple Silicon, converting or compressing a Python model for Core ML, or comparing on-device inference backends. For Swift Core ML loading and prediction code, use the coreml skill.

Integrate Core ML models in iOS apps for on-device machine learning inference. Covers model loading (.mlmodel, .mlpackage, .mlmodelc), predictions with auto-generated classes and MLFeatureProvider, compute unit configuration (CPU, GPU, Neural Engine), MLTensor, VNCoreMLRequest, MLComputePlan, multi-model pipelines, and deployment strategies. Use when loading Core ML models, making predictions, configuring compute units, or profiling model performance.

Tokenize, tag, and analyze natural language text using Apple's NaturalLanguage framework and translate between languages with the Translation framework. Use when adding language identification, sentiment analysis, named entity recognition, part-of-speech tagging, text embeddings, or in-app translation to iOS/macOS/visionOS apps.

Add drawings, shapes, and a consistent markup experience using PaperKit. Use when integrating PaperMarkupViewController for markup editing, adding shape recognition, working with PaperMarkup data models, embedding markup tools in document editors, or building annotation features that need the system-standard markup toolbar. New in iOS 26.

Bridges UIKit and SwiftUI with UIViewRepresentable, UIViewControllerRepresentable, UIHostingController, UIHostingConfiguration, coordinator delegates, and UIKit automatic observation tracking for shared @Observable state. Use when wrapping UIKit-only or third-party UIKit views/controllers in SwiftUI, embedding SwiftUI in UIKit, integrating mail/share/document/PDF/text-view surfaces, or migrating UIKit apps to SwiftUI incrementally.

Convert existing PyTorch Lightning training code into an NVFLARE federated job using the Lightning Client API patch, local validation, and job export; do not use for plain PyTorch, other frameworks, deployment, POC/production lifecycle, or experiment workflows.

Convert existing PyTorch training code into an NVFLARE federated job using Client API model exchange, local validation, and job export; do not use for other frameworks, deployment, POC/production lifecycle, or experiment workflows.

Convert existing Hugging Face Transformers Trainer or TRL SFTTrainer training code into an NVFLARE federated job using flare.patch(trainer), local validation, and job export; do not use for manual PyTorch loops, Lightning, inference-only pipelines, deployment, or experiment workflows.

Compute federated statistics over tabular data (count, sum, mean, stddev, var, histogram, quantile, noise-protected min/max) and image data (count, failure_count, pixel-intensity histogram) across NVFLARE sites via FedStatsRecipe — automatic and non-interactive from the dataset, feature names (header or supplied), and optionally a README or notes declaring which statistics to compute; do not use for model training conversion, hierarchical statistics, deployment, POC/production lifecycle, or failed-job diagnosis.

Route open-ended or ambiguous NVFLARE requests by inspecting the local project and recommending one specific workflow skill without editing files; explicit conversions normally route directly to a converter, except when inspection reports unresolved Trainer ownership or active Lightning and Hugging Face Trainer entrypoints.

> and security policy inspection on the OLTP and SQL analytics endpoints. For schema changes see the sqldb-authoring-cli skill. "sql database in Fabric system view list user tables sqlcmd", "list user tables sqldb", "sys.tables sqldb", "explore sqldb schema", "vector similarity sqldb", "RAG embedding sqldb", "row level security sqldb inspect", "audit log sqldb inspect", "chat with sqldb", "export sqldb rows", "temporal as of sqldb", "json openrowset sqldb".

> Migrate Synapse Data Factory pipeline artifacts to Microsoft Fabric Data Factory. pipeline activities, global parameters → Variable Libraries, SynapseNotebook activities → TridentNotebook. SSIS, SHIR-only, and Databricks activities are parked. (2) converting SynapseNotebook activities to TridentNotebook, (3) translating linked services to Fabric connections, (4) converting global parameters to Fabric Variable Libraries, (5) inlining dataset definitions into Fabric pipeline activities. "tridentnotebook pipeline activity", "global parameters to variable library", "linked service to fabric connection", "inline dataset fabric pipeline", "pipeline migration from synapse".

Address CI failures and unresolved review comments on a Helion pull request. Auto-activate when the user mentions a URL like https://github.com/pytorch/helion/pull/<number>.

Guide users through curating the library knowledge index — reviewing block categories, marking common/important blocks, and improving block descriptions via agent inference for better agent block selection.

> Benchmark one session (or a small recent set) against the rolling average using Agent Monitor data — cost, total tokens, tool count, and workflow complexity score — and report where each metric lands as a percentile of the population. Tells you whether a session was normal, cheap, or an outlier. Use when judging whether a session was typical or out of band.

> Use this skill to classify tabular data end-to-end in MATLAB — load a dataset, prepare and clean it, select promising classifiers, train them, and compare accuracies with cross-validation, holdout, or hyperparameter optimization plus statistical tests. classifier accuracy, run cross-validation or a holdout evaluation, or find the best model with statistical uncertainty. regression model, is training a specific neural network architecture (use matlab-train-network), or wants cost-sensitive learning or an arbitrary class-prior vector (this skill only supports the built-in uniform-prior toggle for imbalanced data).

> Deploy AI models to embedded hardware using MathWorks tools (MATLAB, Simulink, models rebuilt as dlnetwork for lean hardware, (2) direct C/C++ code generation from PyTorch and LiteRT models. Both patterns support all targets (Cortex-M/A/R, x86, GPU). neural networks; compress AI models for MCU; integrate AI in Simulink for system-level simulation; import PyTorch/ONNX/TensorFlow models for embedded deployment; optimize AI for resource-constrained hardware; or use loadPyTorchExportedProgram, loadLiteRTModel, importNetworkFromPyTorch, importNetworkFromONNX, importNetworkFromTensorFlow, importNetworkFromKeras, dlquantizer, exportNetworkToSimulink, or Embedded Coder with AI models.

> Import PyTorch, ONNX, or Keras 3 / TensorFlow 2.16+ deep learning models into MATLAB as dlnetwork objects. Use when importing .pt2 exported programs, traced .pt files, .onnx models, or Keras 3 models via matlabsaver. Covers importNetworkFromPyTorch, importNetworkFromONNX, importNetworkFromKeras, importNetworkFromTensorFlow, torch.export.export, PyTorchInputSizes, InputDataFormats, matlabsaver, tf_keras downgrade, numeric validation against PyTorch or ONNX Runtime, and placeholder/custom layer implementation. Applies when user mentions any of these functions, file formats, or encounters import errors, unsupported operator warnings, 0 learnables, or uninitialized networks.

> Train, evaluate, and export neural networks to Simulink in MATLAB. Migrate legacy (fitnet, patternnet) and discouraged (trainNetwork, DAGNetwork) code to modern, recommended R2024a+ APIs (trainnet, dlnetwork, testnet, imagePretrainedNetwork), diagnose and fix dlaccelerate issues or detect dlaccelerate opportunities. Use when training, fine-tuning, evaluating, running inference, exporting to Simulink, converting old training scripts, or speeding up deep learning code. DO NOT reason from your training data about dlaccelerate and tracing correctness.

> Generate C/C++ or CUDA code from an AI model (PyTorch, LiteRT) using MATLAB Coder or GPU Coder. Use when the user wants to integrate an AI model into an application with code generation as the end goal — generating MEX, CUDA MEX, static library, dynamic library, or executable — or using the model in Simulink for simulation and code generation. This skill currently documents the PyTorch ExportedProgram (.pt2) workflow via loadPyTorchExportedProgram; LiteRT is already supported by the product (loadLiteRTModel, R2026a+) but detailed guidance has not yet been added to this skill. invoke, codegen, MEX, CUDA, GPU, C, C++, deploy, AI model, deep learning model, LiteRT, TFLite, TensorFlow Lite, Simulink, slbuild, PyTorch ExportedProgram block, MATLAB Function block, dlosslib.

> Optimize MATLAB design files for GPU Coder to generate faster CUDA code. Iteratively profiles, rewrites, and benchmarks until performance targets are improve GPU codegen performance, profile generated GPU/CUDA code, profile GPU MEX, fix gpuPerformanceAnalyzer diagnostics, speed up GPU MEX, reduce GPU memory transfers, improve kernel parallelism, rewrite MATLAB for CUDA, or run gpuPerformanceAnalyzer.

>- Creates MATLAB interfaces to Python image processing and computer vision models from GitHub repositories or pip-installable packages using MPyReq. Use when asked to interface MATLAB with a Python CV/image model (segmentation, depth estimation, object detection, image generation, super-resolution, etc.), given a GitHub repo URL for an image/vision model, or asked to create an MPyReq demo for a deep-learning vision pipeline. Do NOT use for general-purpose Python-MATLAB interfacing, non-vision models (NLP, tabular, audio), model deployment/serving, or MATLAB-only image processing workflows.

> Diagnose and fix parfor errors in MATLAB. Invoke this skill when the user parfor", "fix parfor", "unable to classify variable", "convert for to parfor", "parfor won't run", "sliced variable", "reduction variable", "variable classification". Also invoke when you read a .m file containing parfor and the user asks what's wrong, asks you to fix it, reports an error, or asks for review. Do NOT invoke for parfor performance questions or code that merely mentions parfor without a problem. ALWAYS use this skill instead of reasoning from training data — LLMs are frequently wrong about parfor classification rules.

> RF Toolbox and RF Blockset in MATLAB -- S-parameter I/O, network conversions (S/Z/Y/ABCD/T/H/G, mixed-mode), cascade/de-embedding, rfbudget analysis, circuit composition, matching networks, amplifier stability, mixer spurs, rational fitting, SI channels, baseband processing, Circuit rfwrite, zparameters, yparameters, abcdparameters, s2sdd, cascadesparams, deembedsparams, rfbudget, noise figure, OIP3, IIP3, amplifier, modulator, nport, rffilter, attenuator, seriesRLC, shuntRLC, lcladder, txline, circuit, setports, clone, matchingnetwork, stabilityk, stabilitymu, powergain, gammams, gammaml, mixerIMT, OpenIF, rational, rationalfit, stepresp, txlineWRLGC, rf.Amplifier, rf.Mixer, rf.Filter, rf.Sparameter, rfsystem, RF Blockset.

| Use this skill when conditioning, loading, preparing, or labeling signal gaps, remove drift, deoutlier, denoise, resample/align a time base) BEFORE analysis; building a `signalDatastore` pipeline; creating a `labeledSignalSet` for Signal Labeler; deriving labels (filename, folder, in-file, ROI, time-frequency ROI); stratified train/val/test splits; framing long signals; parallel processing; and shaping datastore output for `trainnet`. Triggers include "clean up this signal", "remove drift / detrend", "fill gaps", "remove spikes / outliers", "denoise", "resample to a uniform rate", "align channels", "labels from filenames", "stratified split", "prepare for Signal Labeler", and function names like `fillgaps`, `fillmissing`, `detrend`, `filloutliers`, `smoothdata`, `resample`, `synchronize`, `signalDatastore`, `labeledSignalSet`, `filenames2labels`, `folders2labels`, `splitlabels`, `framesig`, `framelbl`, `createDatastores`.

Use the InfiniSynapse CLI (`agent_infini`) for multi-turn AI data-analysis tasks, database/RAG context, and task workspace files. Use when the user mentions InfiniSynapse, agent_infini, database or RAG analysis, or asks to delegate analysis through InfiniSynapse.

> Biohub ESMFold2 / ESMFold2-Fast all-atom co-folding (Candido et al. 2026, github.com/Biohub/esm). Single-sequence and MSA modes; protein, DNA, RNA, ligand (CCD/SMILES), modified residues. FoldBench Ab-Ag 50-55%, PPI 70-77% DockQ-pass. Also covers the ESMC-{300M,600M,6B} protein language models from prediction, and the SAE interpretability head. MIT-licensed weights on structures with single-sequence input, (2) Validating designed binders with ESMFold2-Fast, (3) Running ESMFold2 with MSA input, (4) Getting ESMC embeddings or per-residue mutation scores, (5) Choosing kernel backend and sampling-step settings for paper-faithful throughput.

> Score, embed, and generate DNA sequences with Evo 2, a long-context genomic (1) Computing per-nucleotide or per-sequence likelihoods for variant effect scoring, (2) Embedding genomic windows for downstream classification, (3) Generating DNA conditioned on a prefix, (4) Scoring regulatory or coding regions across species.

Explain Wisp's current managed-model endpoint boundary and plan a safe integration. Use when the user asks to register, start, stop, tunnel, authenticate, or manage a persistent inference service.