What does this skill do, and when should you use it?
This language-agnostic router skill is the Microsoft Foundry entry point in the Microsoft Agent Skills repository. It maps requests to specialized workflows for agent development, infrastructure, model deployment, evaluation, monitoring, permissions, networking, troubleshooting, and fine-tuning. It requires Azure tooling, Foundry MCP access, and local project context such as azure.yaml or .foundry metadata for many workflows. It is a strong fit for teams operating Microsoft Foundry workloads, but not a general Azure deployment or Azure Functions skill.
Reads azure.yaml, azd environment values, .foundry metadata, and eval.yaml when configuration context is needed; discovers available Foundry MCP operations through the Azure MCP foundry tool; routes requests to sub-skills for deployment, invocation, evaluation, tracing, troubleshooting, project and resource creation, model deployment, quota, RBAC, private networking, or fine-tuning; uses or guides azd, Azure CLI, and Foundry MCP workflows for agent scaffolding, deployment, invocation, evaluation, continuous monitoring, prompt optimization, model customization, and resource configuration; and maintains applicable .foundry metadata, dataset, evaluator, suite, and result caches.
- An Azure developer needs to scaffold, deploy, and smoke-test a new Microsoft Foundry hosted agent.
- An engineering team has existing agent code and needs to deploy or redeploy it with azd and verify the resulting version.
- An AI team wants batch evaluation, continuous production evaluation, prompt optimization, or evaluation datasets generated from production traces.
- A platform engineer needs model deployment, regional capacity or quota guidance, or help diagnosing a Foundry deployment failure.
- An Azure administrator needs Foundry project setup, RBAC, managed identity, service-principal, or network-isolation guidance.
- A model engineer needs SFT, DPO, or RFT workflows covering training data, graders, checkpoints, deployment, and evaluation.
How do you install this skill?
- This assessment is static only; no MCP tools, azd, Azure CLI, scripts, or deployment workflows were executed.
- Before use, verify every linked sub-skill, script, MCP parameter, permission scope, cost impact, data residency requirement, and rollback plan.
- The fine-tuning documentation contains version 0.0.0-placeholder, and some capabilities may require preview access or feature flags; do not treat examples as production validation.
- No evidence was supplied for Chinese-language support or reachability from mainland-China networks.
- Shell / CLI
- Network access
- Local filesystem
- MCP Server
Azure CLIAzure Developer CLI (azd)Azure MCPFoundry MCP
From the target Agent configuration environment, run npx skills add microsoft/skills and select the needed skills in the wizard. Skills are installed into the selected Agent directory, such as .github/skills/ for GitHub Copilot. The collection can also be installed by copying a specific skill directory or by installing the azure-skills plugin through Copilot CLI. The supplied source does not document a separate installer command for selecting only .github/plugins/azure-skills/skills/microsoft-foundry/.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/microsoft/skills.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/.github/plugins/azure-skills/skills/microsoft-foundry" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Once installed, send your agent any of these to trigger it:
- create and deploy a Microsoft Foundry hosted agent
- evaluate and optimize this Foundry agent’s prompt.
In a coding Agent with Azure, azd, Foundry MCP, and the target project configured, make a concrete request such as “create and deploy a Microsoft Foundry hosted agent” or “evaluate and optimize this Foundry agent’s prompt.” The skill first discovers Foundry MCP tools, then requires the matching workflow document to be read before execution. The supplied source does not provide complete commands, credential setup, or parameter examples for every sub-skill.
What are this skill's strengths and limitations?
- Covers the Foundry agent lifecycle from creation and deployment through invocation, evaluation, optimization, and troubleshooting.
- Includes infrastructure and operations topics such as model deployment, quota, RBAC, private networking, and fine-tuning.
- Provides explicit intent-routing rules, anti-triggers, and workflow-specific pre-execution requirements.
- MIT licensed, with repository-level testing and evaluation automation described in the README.
- It is a router skill, so many concrete procedures depend on sub-skill documents not included in the supplied material.
- It explicitly depends on Azure CLI, azd, MCP, and Foundry project context, so it is not a general-purpose agent skill for non-Azure environments.
- The source does not establish that every platform, region, model, or preview feature is validated.
- The README’s test-coverage summary describes the repository collection and does not prove independent coverage for this specific skill.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Microsoft Foundry Agent Skill this page ✓ Microsoft · Official | 47 · Use with care | ★ 2.7k | 3mo ago | MIT |
| Azure Skill Creator ✓ Microsoft · Official | 48 · Use with care | ★ 2.7k | 3mo ago | MIT |
| Microsoft Agent Skills ✓ Microsoft · Official | Not yet reviewed | ★ 2.7k | 3mo ago | MIT |
| MCP Server Development Guide ✓ Microsoft · Official | 49 · Use with care | ★ 2.7k | 3mo ago | MIT |
| Microsoft Foundry Python Projects Skill ✓ Microsoft · Official | 47 · Use with care | ★ 2.7k | 3mo ago | MIT |
How did FollowSkills review this skill?
The evidence requires Foundry MCP discovery, mandatory workflow-document reads, scoped workspace resolution, and user prompts for missing configuration. It does not clearly specify per-operation permissions, sensitive-data handling, confirmation of external effects, rollback, or complete data-flow disclosure, so points are deducted. No red-line risk is evident.
The evidence provides workflow routing, prechecks, context-resolution rules, and some failure guidance. However, key sub-skills, scripts, and MCP parameters are not fully supplied, and static review cannot reproduce deployment, evaluation, or troubleshooting paths, so points are deducted.
USE FOR/DO NOT USE FOR boundaries, workflow tables, and multiple Foundry scenarios provide a reasonably clear audience and trigger surface. The scope is very broad, trigger precision and non-fit limits depend on unavailable sub-skills, and there is no evidence for Chinese support or mainland-China network reachability, so points are deducted.
The evidence includes YAML metadata, MIT licensing, a version, layered sub-skills, installation guidance, examples, and limitation notes. The finetuning sub-skill has version 0.0.0-placeholder, while changelog, maintenance ownership, update path, and documentation consistency are incomplete, so points are deducted.
The evidence covers core lifecycle tasks from resource creation through deployment, invocation, evaluation, fine-tuning, and troubleshooting, with commands, formats, and error guidance. No directly usable output or execution result for this skill's key paths is supplied, so environment adaptation and review burden remain material.
The repository context includes real CI workflows and test code, but the supplied material does not establish coverage of this microsoft-foundry skill's key paths and provides no third-party execution results or independent corroboration. Static calibration therefore supports only limited credit.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →