83 skills in force: 27 Google, 27 Microsoft, 21 Hugging Face, 5 OpenAI, 1 Firebase (Google), 1 Anthropic, 1 Sentry. What is a skill?
developing-genkit-dart
Genkit Dart is an AI SDK for Dart and Flutter: generation, structured outputs, tools, streams, agents, prompts in Dotprompt format, and a generative interface (A2UI).
developing-genkit-go
Genkit for Go Unifies Generation, Structured Output, Continuous Flow, Tool Call, Invites and Streams, regardless of the model provider (the SKILL.md relies on Firebase/genkit.go packages and Google GenAI plugin).
developing-genkit-js
Genkit guide for Node.js and TypeScript: feed, tools, Dotprompt prompts, observability, middleware, and troubleshooting errors, validation or typing.
developing-genkit-python
This skill target Genkit Python 0.13: generation, structured output, streams, tools, multi-tour agents and FastAPI routes, as well as migration since 0.12 (cracked imports and signatures).
gemini-interactions-api
The API Interactions, on Gemini Enterprise Agent Platform, is the recommended interface for multi-turn conversations, background tasks, stream, structured output and function call, with server-side managed status.
pathways-workload
Debugging a Pathways on Cloud load — JAX training or inference on TPU Cloud in GKE, in the form of JobSet, PathwaysJob or LeaderWorkerSet — that fails, crashes, restarts or no longer progresses.
azure-ai
This skill guides the construction and troubleshooting of Azure AI Search, Speech, OpenAI and Document Intelligence applications.
gemini-live-api
This skill generates, in the chosen language, a client class for API Live Gemini Enterprise: WebSocket connection to the endpoint, session (establishment and resumption), refreshment of the carrier token via default application identifiers, and exchange of CustomerMessage / ServerMessage messages.
gemini-api
It is used as soon as the application relates to Gemini in a company context, or explicitly mentions Vertex AI, Google Cloud or Agent Platform — the latter name, Gemini Enterprise Agent Platform, replaces the old Vertex AI brand still common in web documentation.
agent-platform-architecture
Design and validation guide for a corporate Gemini architecture: multi-project hierarchy (host, central governance hub, execution radius, tool radius or MCP), Agent Gateway, connectivity models (all traffic or only private beaches), secure web proxy with policy routing or inter-cloud mTLS output hub, private certification authority and internal balancer, unified access policies (CEL) and semantic governance, third party identity provider (Entra ID, Ping, Okta) or SPIFFE agent identity management, Model Armor safeguards, inter-project registry synchronization, ADK agents on the runtime, and day operations 2.
foundry-iq
Covers the knowledge bases Foundry IQ: make local or Blob documents searchable, create or diagnose a database, sort an unsupported connector or multi-source database (even with a "classic" search vocabulary), connect an existing database to agents, create or reuse a Search service, and query a database with citations.
agent-platform-troubleshooting
Diagnostic Runbook, no automatic repair: the user keeps his hand on any mutation.
agent-platform-deploy
It is used to deploy an open model or custom weights from Model Garden to an Agent Platform endpoint, to copy and deploy an already refined first-party model, to list the deployable models catalogue, check a deployment configuration, estimate a cost, track an ongoing operation, troubleshoot an error (e.g.
agent-platform-eval-flywheel
This skill measures and improves the quality of a model or agent on Google Cloud, using the Eval Quality Flywheel method.
agent-platform-tuning
This skill leads a fine tuning of Gemini models or open models on the Agent Platform infrastructure, from environmental preparation to monitoring, data set and job configuration.
agent-platform-rag-engine-management
This skill drives the RAG engine of Agent Platform: listing or inspecting corpus and files, recovering contexts, or generating content anchored in a corpus, via the SDK Python `vertexai` (Google GenAI), and not by raw REST calls or MCP tools, because the code is intended for external customers.
agent-platform-model-registry
The Agent Platform register keeps the learning models and their versions.
agent-platform-endpoint-management
An Agent Platform endpoint is a logical service host, with a stable URL for online predictions: it must be created before deploying a model.
agent-platform-prompt-management
Here we create, list, retrieve, release or delete prompts managed in Agent Platform, typically via SDK and Python extracts executed in the environment, once the operation is approved.
agent-platform-inference
This skill is used to authenticate and interrogate models of Agent Platform: Gemini first-party models, third-party open models (OpenMaaS: Llama, DeepSeek, Qwen, etc.), or a custom endpoint identified by a resource number (Gemini refined model, LLM deployed from Model Garden, historical custom model).
agent-platform-tuning-management
Unlike the skill that launches a fine-tuning, it manages existing jobs via the SDK Python ofAgent Platform: listing performances, getting details, or canceling a job that runs too long.
firebase-ai-logic-basics
Firebase AI Logic (formerly Vertex AI in Firebase) makes it possible to call Gemini models from the client code of a mobile or web app (Android, iOS, Web, Flutter): text generation, multimodal input, chat, streaming, structured output, hybrid inference on the device, App Check protection, and model switch via Remote Config.
hf-cli
Learns to the agent all the commands of the `hf` CLI: download and publish models, dataset and Spaces, manage cache, fuckets, repositories, jobs, articles, webhooks and Inference Endpoints.
microsoft-foundry
Covers the full life cycle of Microsoft Foundry agents, models and resources: construction, deployment, evaluation, optimization, refining and operation, including azd, hosted agents, tools, batch or continuous evaluations, IC/CD, agent optimizer, model deployment and refining (SFT, DPO, RFT), RBAC, quotas, regions and knowledge indexes.
apigee-agent-gateway-governance
This skill sets up an Apigee proxy that an Agent Gateway Google Cloud calls in ext-proc (Service Extension), on the model traffic of an agent.
agent-platform-migrate-from-ai-studio
The transition takes place from an application centered on the API Gemini d-AI Studio (generativelanguage.googleapis.com) to the corporate agent platform on Google Cloud (aiplatform.googleapis.com, formerly in the Vertex AI orbit).
google-cloud-solution-guided-gke-ai-migration
Migration guide for an existing IA inference — Cloud Run, API Gemini, Gemini Enterprise Agent Platform, or VM custom — to a self-hosted inference on GKE, with gcloud and kubectl.
google-cloud-solution-rag-enterprise-search-gke-sqldb
Design of a business search in generation augmented by recovery (RAG), where the store and vector index are a vector-capable SQL database (the skill includes AlloyDB), the model and the inference frame are open, and all the application components rotate in Kubernetes containers, typically on GKE.
google-cloud-solution-agentic-ai-bidirectional-streaming
Interactive discovery of needs for a living multi-agent system, which continuously deals with multimodal flows, typically for real-time technical guidance or safety supervision.
gke-ai-troubleshooting-tpu-mxla-hang
A multi-slice TPU drive that is no longer progressing, writes HANG DETECTED, or remains blocked in a collective operation, is not the same as a simple drop in flow.
gke-ai-troubleshooting-tpu-performance-degradation
Here, the pods still rotate, but the drive rate drops and the time per step decreases: typically a 15% or more drop in the duty cycle TPU, possibly reported by a PERFORMANCE DEGRADATION event.
gke-inference
This skill deploys and adjusts IA inference loads on GKE, using GPU, TPU and model servers, relying on Google Inference Quickstart and LLM service practices.
azure-search-nav
This skill is about building a deep link to a blade or menu of the Azure portal, for Microsoft resources.ContainerService, Microsoft.Kubernets (including Arc) and Microsoft.Compute.
claude-api
To be used as an up-to-date reference for building an application that calls the API Claude or an SDK Anthropic: model identifiers, parameters, streaming, tool call, MCP, agents, cache, token count and migration between models.
trl-training
The post-training skills train LLMs with TRL (Transformers Reinforcement Learning): SFT, DPO, GRPO, KTO and reward model training, whether through the API Python or the CLI trl.
preset
Deploys an Azure OpenAI model to an optimal region, analysing the capacity on all available regions: the current region is checked first, then alternatives are proposed if necessary.
deploy-model
Unified deployment skills of Azure OpenAI models, which are directed according to the intent: preset rapi deployment, fully customized deployment (version, SKU, capability, RAI policy) or capacity discovery between regions and projects.
hf-cloud-serving-image-selection
The skill selects the right service container for a SageMaker deployment and finds the image to be up-to-date, as soon as a termination point is set and an image has to be selected (LLM, model Hugging Face, fine tune, embeddings, reranker, sentence-transformers), or before freezing an URI in the code.
hf-cloud-sagemaker-production-defaults
Last step of the SageMaker deployment stream: create a termination point (real time, real time scale-to-zero, or asynchronous) with default autoscaling, CloudWatch alarms and tags.
train-sentence-transformers
The skill drives or refines sentence-transformer models: SentenceTransformer (bi-encoder, dense or static, for search, similarity, clustering, classification, paraphrase, deduplication, multimodal), CrossEncoder (reranker, pair score), SpraseEncoder (SPLADE, hollow lives) and MultiVectorEncoder (ColBERT / late interaction, MaxSim).
capacity
Discover the available capacity of Azure OpenAI models, between regions and projects: quota limit analysis, availability comparison (including TPM) and deployment site recommendation.
finetuning
Guide to model refining on Microsoft Foundry, in SFT (supervised), DPO (preferences) or RFT (reinforcement with evaluators): preparation of the data set, submission of training work, deployment and evaluation, including distillation, calibration of an evaluator and uploading of large files.
customize
Offers a guided, interactive stream to deploy an Azure OpenAI model with complete control: version, SKU (GlobalStandard, Standard or ProvidedManaged), capability, RAI policy (content filter) and advanced options (dynamic quota, priority, spillover, forecasted flow).
huggingface-spaces
The skill builds, deploys and maintains an application on Hugging Face Spaces: SDK Gradio, Docker or static, ZeroGPU or dedicated hardware, model loading, debugging, fuckets, inference providers, community grants.
hf-cloud-sagemaker-deployment-planner
This skill is the entry point for planning and coordinating the deployment of a model on Amazon SageMaker AI, even if the demand is vague ("turn it on AWS").
huggingface-best
The skill helps to find the best model, or a benchmark comparison, for a given task, including "what model runs on my laptop".
azure-ai-language-conversations-py
This skill guides the integration of Conversational Language Understanding (CLU) with the SDK Python `azure-ai-language-conversations`.
azure-aigateway
Guides the configuration of API Management as a gateway for models, MCP tools and agents: semantic cache, token limit, content security, load distribution, model governance, MCP flow limitation, jailbreak detection, addition of Azure OpenAI backend or Foundry model, LLM policies, metrics and cost control, or conversion of API to MCP (OpenAPI import).
azure-speech-to-text-rest-py
API REST Azure Speech to Text for short audio, in Python, intended for simple recognition of files of up to 60 seconds, without the Speech SDK.
azure-ai-textanalytics-py
This skill guides the SDK Python Azure AI Text Analytics (`azure-ai-textanalytics`) and `TextAnalyticsClient`.
azure-ai-vision-imageanalysis-py
This skill covers the SDK Python Azure AI Vision Image Analysis (`azure-ai-vision-imageanalysis`) and the `ImageAnalysisClient` client.
azure-ai-voicelive-py
SDK Azure AI Voice Live for Python (azure-ai-voicelive), intended for applications that require real-time two-way audio communication with Azure AI: voice assistants, voice chatbots, live speech translation, voice-driven avatars, or any WebSocket audio stream with AI models.
azure-ai-ml-py
The SDK Azure Machine Learning v2 for Python controls workspaces, jobs, models, datasets, calculation and pipelines, mainly via MLClient.
azure-ai-transcription-py
The SDK Azure AI Transcription for Python converts speech into text, in real time or in batches, with time stamping and diarization, around TranscriptClient.
huggingface-llm-trainer
Trains or refines language and vision models with TRL or Unsloth on Hugging Face infrastructure Jobs: SFT, DPO, GRPO, reward templates, GGUF conversion, hardware selection and cost estimation.
huggingface-trackio
Trackio follows and visualizes training experiments: metric logging (API Python), diagnostic alerts, and metric reading or analysis (CLI).
hf-mem
hf-mem is a Hugging Face CLI that estimates the memory needed to load Safetensors or GGUF weights in inference from the Hub.
huggingface-zerogpu
The skills frame demos and GPU computing on ZeroGPU, in Gradio Spaces and Hugging Face Spaces.
scribe
Serves when using Scribe, the transcription service of the Zoom IA services.
azure-search-documents-py
The SDK Azure AI Search for Python is used for vector research, hybrid or semantic, indexing and skills.
conversation-intelligence
Twilio Conversation Intelligence Development Guide.
twilio-voice-conversation-relay
To build voice agents assisted by the AI, with Twilio ConversationRelay.
openai-api-troubleshooting
To classify the probable cause of an already sent OpenAI request that failed, explain the next step and return to the adapted skill.
ai-gateway
Expert guide on Vercel AI Gateway.
azure-ai-voicelive-dotnet
This skill is about the SDK .NET Azure AI Voice Live (`Azure.AI.VoiceLive`), for real-time voice applications with two-way WebSocket communication.
azure-search-documents-dotnet
The source text describes the SDK .NET Azure AI Search (`Azure.Search.Documents`).
prompt-optimizer
Serves to create, optimize and refine iteratively fast agents, systems, developers and reusable models, including to carry them between OpenAI, Claude or Gemini, or to build prompt evaluations.
azure-mgmt-weightsandbiases-dotnet
This skill covers the Weights & Biases control plan on Azure, in .NET.
azure-ai-openai-dotnet
This skill covers the SDK .NET Azure OpenAI (`Azure.AI.OpenAI`), client library for Azure OpenAI and OpenAI services: chat complements, dives, image generation, audio transcription, and assistants.
azure-ai-vision-imageanalysis-java
The SDK Azure AI Vision Image Analysis for Java builds image analysis applications: legend, OCR, object detection, labelling and intelligent cropping.
azure-ai-anomalydetector-java
The SDK Azure AI Anomaly Detector for Java detects univariate or multivariate anomalies, analyses time series and feeds assisted supervision.
azure-ai-voicelive-java
The SDK Azure AI Voice Live for Java is conducting a real-time two-way voice conversation with a WebSocket assistant around VoiceLiveClient and a VoiceLiveSessionAsyncClient for the streaming session.
huggingface-local-models
The skill helps to choose and run a model locally with llama.cpp and GGUF, on CPU, Mac Metal, CUDA or ROCm.
azure-ai-voicelive-ts
This skill is about the SDK JavaScript Azure AI Voice Live (`@azure/ai-voicelive` and `VoiceLiveClient`), for real-time voice applications with two-way WebSocket, Node.js side or browser.
azure-search-documents-ts
This skill is used for the SDK JavaScript Azure AI Search (`@azure/search-documents`): create or manage indexes, and set up vector or hybrid search, typeScript side.
transformers-js
The skill uses Transformers.js to run machine learning models directly in JavaScript or TypeScript, in the browser or on the server side (Node.js, Bun, Deno), via WebGPU or WASM, from pre-trained Hub models.
huggingface-tool-builder
The skill builds a reusable script to search, enrich or process data via the API Hugging Face, especially when multiple calls have to be chained or a task will be repeated.
huggingface-gradio
The skill helps build Gradio Python interfaces and demos: applications, components, event headphones, layouts and chatbots.
huggingface-vision-trainer
Training and refining vision models (object detection, image classification, SAM/SAM2 segmentation) with Transformers on cloud GPUs from Hugging Face Jobs: COCO data preparation, increase, evaluation.
huggingface-community-evals
The skill launches model evaluations of the Hub with inspection-ai and lighteval, on local equipment.
huggingface-papers
The skills search and read the articles pages Hugging Face in Markdown, and query the API papers for structured metadata (authors, models, dataset, related Spaces, GitHub repository, project page).
huggingface-paper-publisher
The skill publishes and manages research articles on the Hugging Face Hub: article pages, links to models or datasets, claims from the water, and generation of professional articles in Markdown.
hf-mcp
To use the Hugging Face Hub through the MCP server tools: to search for models, data sets, Spaces and articles, consult a repository or documentation, start computing tasks, and use Spaces Gradio as the tools of the AIA.
Skills from the openai/skills repository, impaired to openai/plugins.
openai-docs
Answers questions about construction with products and OpenAI APIs.