Azure W&B Resource Manager for .NET
Deploy and manage Weights & Biases instances in Azure with the .NET ARM SDK.
The skill uses Azure identity authentication and mentions managed identity and SSO; no malware or covert exfiltration is evident. However, its examples create, update, and delete cloud resources, involve Marketplace subscriptions and administrator email data, and omit user confirmation, least-privilege guidance, sensitive-data handling, rollback, and deletion recovery, so points are deducted.
Installation, authentication, resource-management, and error-handling paths are documented, and the repository contains generic CI and tests. Static evidence also shows a preview API, no skill-specific tests, synchronous .Result mixed into async code, and a name-availability check based on indirect 404 inference; abnormal-input and failure-diagnosis coverage is thin, so the score is capped and reduced.
Triggers, target SDK, core workflows, and .NET environment are reasonably clear. The skill does not define non-fit boundaries, input/output contracts, permission prerequisites, preview or regional constraints, mainland-China reachability, or Chinese-language support, so points are deducted.
The document has clear sections for installation, environment variables, versions/API version, types, regions, best practices, errors, related SDKs, MIT licensing, and authorship. It lacks a skill-specific changelog, explicit maintenance owner/update path, known limitations, compatibility guidance, and FAQ, so points are deducted.
It covers creation, retrieval, listing, SSO, updates, deletion, name checks, and W&B SDK integration, giving broad task coverage. Potential API/compilation and semantic issues, preview status, and lack of skill-specific validation mean generated output still requires substantial review and correction; the static cap also applies.
The skill lists NuGet, documentation, Marketplace, and GitHub references, while the repository includes generic workflows and tests. There is no skill-specific acceptance-criteria suite, CI coverage, or third-party execution evidence, so key claims are not independently reproducible from the supplied static material and only limited credit is warranted.
- Creating and deleting Azure Marketplace resources can cause external effects and charges; explicit confirmation, permission, billing, data-location, and recovery guidance should be added before execution.
- The examples use a preview package/API, and some call chains and the name-availability logic may require correction against the actual SDK signatures; compile validation against the target version is needed.
- DefaultAzureCredential may use multiple local or environment credential sources; credential provenance, least-privilege RBAC, and production configuration should be stated.
- The skill does not document mainland-China reachability, regional availability, or Chinese-language support.
What it does & when to use it
This skill helps developers use the Azure Resource Manager .NET SDK to manage Weights & Biases instances deployed through Azure Marketplace. It covers creation, retrieval, listing, updates, deletion, name checks, and Entra ID SSO configuration. The examples include marketplace offer data, administrator details, partner configuration, optional managed identity, tagging, and Azure error handling. The documented package is v1.0.0-beta.1 and targets API version 2024-09-18-preview.
Guides an agent in producing or editing C# code that installs Azure.ResourceManager.WeightsAndBiases and Azure.Identity, creates an ArmClient with DefaultAzureCredential, accesses instances through resource groups or subscriptions, configures marketplace details, administrators, partner properties, managed identity, and SSO, performs create/read/list/patch/delete operations, and handles common 400, 404, and 409 RequestFailedException cases.
- An Azure platform engineer needs to provision a Weights & Biases instance from Azure Marketplace with C#.
- An ML platform team needs to inspect instance location, subdomain, provisioning state, and partner properties.
- An enterprise developer needs to configure SAML-based Entra ID SSO for a W&B instance.
- A cloud administrator needs to apply cost-allocation tags or enumerate W&B instances across resource groups and subscriptions.
- An ML team needs to connect the Azure-deployed instance to the W&B Python SDK for experiment runs and metric logging.
Pros & cons
- Covers the full ARM resource lifecycle: create, retrieve, list, update, and delete.
- Includes C# examples for Marketplace integration, SSO, managed identity, tags, and error handling.
- Documents the preview package version, API version, supported regions, and important resource types.
- Uses the MIT license, and the parent repository is actively maintained with evolving SDK patterns and expanded tests.
- The SDK and API are preview releases; the documented package is v1.0.0-beta.1.
- Examples require an Azure subscription, resource group, Marketplace configuration, and usable credentials.
- The source does not provide independent test coverage or platform validation for this individual skill.
- The name-availability example infers availability from a 404 response and notes that the SDK may not expose a dedicated operation.
How to install
Install the collection with npx skills add microsoft/skills, then select the required skill in the wizard. The README says skills are installed into the selected agent directory, such as .github/skills/ for GitHub Copilot; a dedicated installation command for this individual skill is not documented. For the .NET SDK, run dotnet add package Azure.ResourceManager.WeightsAndBiases --prerelease and dotnet add package Azure.Identity.
How to use
Give the coding agent a request containing a documented trigger, such as: “Create an Azure Weights & Biases instance with Azure.ResourceManager.WeightsAndBiases, including marketplace settings, managed identity, and Entra SSO.” Set AZURE_SUBSCRIPTION_ID, AZURE_RESOURCE_GROUP, and AZURE_WANDB_INSTANCE_NAME; when using DefaultAzureCredential in production, also set AZURE_TOKEN_CREDENTIALS=prod or a specific credential value.