From 9a4c31b03180dfa11552521935b7a574b12abf0d Mon Sep 17 00:00:00 2001 From: Chenyi An Date: Tue, 30 Jun 2026 15:50:46 +0800 Subject: [PATCH] feat: add foundry skill docs --- doc/microsoft-foundry-skill-status.md | 81 +++++++++++++++++++++++++++ 1 file changed, 81 insertions(+) create mode 100644 doc/microsoft-foundry-skill-status.md diff --git a/doc/microsoft-foundry-skill-status.md b/doc/microsoft-foundry-skill-status.md new file mode 100644 index 0000000..ded9409 --- /dev/null +++ b/doc/microsoft-foundry-skill-status.md @@ -0,0 +1,81 @@ +# Microsoft Foundry Skill Status and Benchmark + +Status and benchmark results for the `microsoft-foundry` skill, with prompts you can use to reproduce the same tests. + +The `microsoft-foundry` skill is used by Copilot-assisted agent development in Microsoft Foundry Toolkit. + +**Source code:** [`microsoft/azure-skills/skills/microsoft-foundry`](https://github.com/microsoft/azure-skills/tree/main/skills/microsoft-foundry) + +**Quick links:** [Install the skill](#install-the-skill) | [Current benchmark](#current-benchmark) | [Run the benchmark yourself](#run-the-benchmark-yourself) + +## Install the Skill + +Install from `microsoft/azure-skills`: + +```bash +npx skills add https://github.com/microsoft/azure-skills --skill microsoft-foundry +``` + +The skill installs to: + +```text +~/.agent/skills/microsoft-foundry +``` + +## Verify the Installation + +Start a fresh GitHub Copilot CLI session and ask: + +```text +Use the microsoft-foundry skill. What can you help me do? +``` + +The response should mention Foundry agent workflows such as scaffolding, local testing, deployment, invocation, and evaluation. + +## Current Benchmark + +The Foundry skill performs consistently across different Copilot models. The benchmark tested Claude Opus 4.8, Claude Opus 4.6, Claude Sonnet 4.6, GPT-5.5, GPT-5.4, GPT-5.3-Codex, GPT-5 mini, and MAI-Code-1-Flash. + +Representative results: + +| Copilot model | Time | AI Credits | +| --- | ---: | ---: | +| Sonnet 4.6 | 10 min 30 s | 100 | +| Auto (GPT-5.3-Codex) | 9 min 30 s | 60 | +| Free plan / cheapest (GPT-5 mini) | 11 min 30 s | 20 | + +These results measure time cost and GitHub Copilot AI Credits for this golden path prompt: + +```text +Create a Python hosted agent for B2B customer onboarding and deploy it to a new Foundry project. Use the Responses protocol. After it is done, run in locally to make sure it can run successfully; then deploy it to foundry and ensure it can respond to users correctly +``` + +## Run the Benchmark Yourself + +### Prerequisites + +- Azure subscription +- Azure CLI (`az`) +- Azure Developer CLI (`azd`) +- Python + +### End-to-end Testing + +Use the golden path prompt: + +```text +Create a Python hosted agent for B2B customer onboarding and deploy it to a new Foundry project. Use the Responses protocol. After it is done, run in locally to make sure it can run successfully; then deploy it to foundry and ensure it can respond to users correctly +``` + +### Stage-by-stage Testing + +Use these prompts when you want to isolate one scenario instead of measuring the full end-to-end path. + +| Scenario | Prompt | +| --- | --- | +| 1a Configuration and scaffolding (greenfield) | `Create a Python hosted agent for B2B customer onboarding and set up whatever I need.` | +| 1b Configuration and scaffolding (brownfield) | `I have an existing agent built with the OpenAI SDK. I want to deploy it as a Foundry hosted agent. What do I need to change?` | +| 2 Local testing | `Help me run this agent locally and test it before I deploy.` | +| 3 Deployment | `I'm happy with it locally. Deploy this to Foundry.` | +| 4 Direct Toolbox iteration | `Add a WorkIQ tool to my deployed Foundry agent and redeploy.` | +| 4-iter Casual iteration | `Can my agent also look up emails and meetings from my Outlook to help with onboarding?` |