[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M69BWVGYYVPE3FX80DDQ3ABA":3,"skill-same:01M69BWVGYYVPE3FX80DDQ3ABA":68},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":31,"publishedAt":40,"linkOnly":41,"markdownContent":42,"markdownBody":43,"frontMatter":44,"refName":45,"commitSha":46,"skillDirectory":47,"forksCount":48,"repositoryHtmlUrl":49,"repositoryHomepageUrl":6,"ownerHtmlUrl":50,"ownerBlogUrl":51,"ownerDescription":52,"updatedAt":53,"install":54,"resources":67},"01M69BWVGYYVPE3FX80DDQ3ABA","ill335-hosted-agent",null,"Supports the ILL335 Microsoft Foundry hosted-agent lab. Use for Lab 6, Agent Framework, FoundryChatClient, ResponsesHostServer, direct code deployment, unified azure.yaml, microsoft.foundry, local readiness checks, azd up, azd deploy, agent invocation, evaluation generation, or deployment troubleshooting.","microsoft\u002Ffrontiertour27-ILL335-building-deploying-ai-agents-with-microsoft-foundry","microsoft","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F6154722?v=4","01M698VYDXZ6JW5WH3CRV650R8","Microsoft","01M69BWP7EAS7CPP7Q2V5Y68G0",23,"MIT","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Ffrontiertour27-ILL335-building-deploying-ai-agents-with-microsoft-foundry\u002Fblob\u002Fmain\u002F.github\u002Fskills\u002Fill335-hosted-agent\u002FSKILL.md",[18,23,26],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000011","运维与部署","SCENARIO","使用场景",{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000023","Skill与智能体开发",{"categoryNo":27,"categoryName":28,"taxonomyType":29,"taxonomyTypeDesc":30},"01JBSEEDCAT000000000000030","开发工程师","PROFESSION","适用职业",[32,9,33,34,35,36,37,38,39],"ill335 hosted agent","foundry","microsoft foundry","agent framework","deploy","azure","deployment","code","2026-11-10 10:12:30",false,"---\nname: ill335-hosted-agent\ndescription: Supports the ILL335 Microsoft Foundry hosted-agent lab. Use for Lab 6, Agent Framework, FoundryChatClient, ResponsesHostServer, direct code deployment, unified azure.yaml, microsoft.foundry, local readiness checks, azd up, azd deploy, agent invocation, evaluation generation, or deployment troubleshooting.\nlicense: MIT\n---\n\n# Support the ILL335 Hosted-Agent Lab\n\nFollow the repository's direct-code deployment path. Use the Microsoft Learn [hosted-agent quickstart](https:\u002F\u002Flearn.microsoft.com\u002Fazure\u002Ffoundry\u002Fagents\u002Fquickstarts\u002Fquickstart-hosted-agent?pivots=azd), [`azure.yaml` reference](https:\u002F\u002Flearn.microsoft.com\u002Fazure\u002Ffoundry\u002Fagents\u002Fconcepts\u002Fazure-yaml-reference), [`azd` development guide](https:\u002F\u002Flearn.microsoft.com\u002Fazure\u002Ffoundry\u002Fagents\u002Fconcepts\u002Fcli-agent-development), and current [`microsoft-foundry\u002Ffoundry-samples`](https:\u002F\u002Fgithub.com\u002Fmicrosoft-foundry\u002Ffoundry-samples\u002Ftree\u002Fmain\u002Fsamples\u002Fpython\u002Fhosted-agents) as the sources of truth.\n\nDo not use older `Azure-Samples\u002Fazd-ai-starter-*` repositories as the reference implementation. Do not add deprecated standalone `agent.yaml` or `agent.manifest.yaml` files; the unified root `azure.yaml` replaces them.\n\nThis lab uses direct-code remote build through `codeConfiguration`. Do not add a Dockerfile, container registry workflow, or capability-host resources unless the lab deliberately changes to a container deployment or Standard Agent Setup.\n\n## Inspect Before Changing\n\nRead:\n\n- `src\u002Fagent\u002Fapp.py`\n- `src\u002Fagent\u002Frequirements.txt`\n- `azure.yaml`\n- `docs\u002Flab6-deploy-agent.md`\n- `.vscode\u002Ftasks.json`\n\nConfirm `azure.yaml` declares:\n\n- `azd` 1.27.1 or later\n- `azure.ai.agents` 1.0.0-beta.8 or later and `azure.ai.projects` 1.0.0-beta.4 or later under `requiredVersions`\n- an `azure.ai.project` service with the model deployment\n- `host: azure.ai.agent`\n- `uses` linking the agent to the project service\n- Python 3.13 under `codeConfiguration`\n- `app.py` as the entry point\n- Responses protocol `2.0.0`\n- `infra.provider: microsoft.foundry`\n\n## Preserve the Current Agent Pattern\n\n```python\nfrom agent_framework import Agent\nfrom agent_framework.foundry import FoundryChatClient\nfrom agent_framework_foundry_hosting import ResponsesHostServer\n\nagent = Agent(\n    client=FoundryChatClient(...),\n    instructions=SYSTEM_PROMPT,\n    default_options={\"store\": False},\n)\n\nResponsesHostServer(agent).run()\n```\n\nFoundry manages conversation continuity. Do not add local conversation-history persistence.\n\n## Prepare the Learner Environment\n\n1. Confirm `azd version` is 1.27.1 or later.\n2. Run `azd ext install microsoft.foundry`. This meta-extension installs the `azure.ai.*` providers; it does not replace their compatible `requiredVersions` entries in `azure.yaml`.\n3. Create and activate `.venv`.\n4. Install `src\u002Fagent\u002Frequirements.txt`.\n5. Run `python -m pip check`.\n6. Confirm Azure CLI and `azd` use the same tenant and subscription before cloud operations.\n7. Never print `.env`, access tokens, or connection strings.\n\n## Validate Locally\n\nUse the Foundry Toolkit F5 task when available. The local server is ready when:\n\n```text\nGET http:\u002F\u002F127.0.0.1:8088\u002Freadiness\n```\n\nreturns HTTP 200 with a healthy status.\n\n## Provision, Deploy, and Invoke\n\nThe first cloud operation runs both provider phases: `azd provision` creates the project and model deployment, then `azd deploy` publishes the hosted-agent version:\n\n```powershell\nazd up\n```\n\nThis operation can create billable Azure resources. State that before running it and require the learner's approval.\n\nFor later code-only updates to an environment that this project already provisioned:\n\n```powershell\nazd deploy\nazd ai agent show --output json\nazd ai agent invoke \"The delivery was late, but support kept me informed.\"\n```\n\nRemote deployment and invocation can incur Azure charges. State that before executing them. Do not run `azd up`, `azd deploy`, or `azd down` merely to test a theory.\n\n## Generate and Run Evaluations\n\nTreat evaluation as an optional post-deployment extension. Generation and evaluation are preview operations that make billable model calls, so state that before running them.\n\nUse `src\u002Fagent\u002Fevaluation-instructions.md` to generate a domain-specific synthetic suite. Keep `src\u002Fagent\u002Fevals\u002Fcaldova-golden.jsonl` as the curated regression set for regulated-routing expectations. Generated data broadens coverage; it does not replace human-reviewed golden cases.\n\n```powershell\nazd ai agent eval generate --agent caldova-consumer-sentiment-agent --gen-instruction-file src\u002Fagent\u002Fevaluation-instructions.md --eval-model gpt-5.4-mini --max-samples 15 --out-file eval.yaml\nazd ai agent eval run --config eval.yaml\n```\n\nInspect generated datasets, evaluators, and `eval.yaml` before running them. Do not run optimizer workflows automatically.\n\nWhen a schema criterion fails, inspect `sample.output_text` before changing the agent. Preview evaluators can mistakenly grade the serialized Responses transport envelope in `sample.output` or `sample.output_items`. If the assistant text is valid but the explanation cites wrapper fields such as annotations or output items, refine the evaluator to grade `sample.output_text`, upload it with `azd ai agent eval update --config eval.yaml --evaluator-only`, and rerun it. Do not present transport-envelope failures as agent regressions.\n\nIf wrapper-based failures persist or valid `sample.output_text` is marked not applicable, report the schema criterion separately as a preview evaluator limitation. Do not use the aggregate pass rate alone as the agent quality result.\n\n## Success Criteria\n\n- Agent dependencies resolve without conflicts.\n- The local readiness endpoint is healthy.\n- `azd up` provisions the project and creates an active hosted-agent version on first use.\n- `azd deploy` updates an active hosted-agent version without reprovisioning infrastructure.\n- Invocation returns governed JSON through the Responses protocol.\n- A learner can distinguish smoke invocation, generated evaluation coverage, and curated regression coverage.\n","# Support the ILL335 Hosted-Agent Lab\n\nFollow the repository's direct-code deployment path. Use the Microsoft Learn [hosted-agent quickstart](https:\u002F\u002Flearn.microsoft.com\u002Fazure\u002Ffoundry\u002Fagents\u002Fquickstarts\u002Fquickstart-hosted-agent?pivots=azd), [`azure.yaml` reference](https:\u002F\u002Flearn.microsoft.com\u002Fazure\u002Ffoundry\u002Fagents\u002Fconcepts\u002Fazure-yaml-reference), [`azd` development guide](https:\u002F\u002Flearn.microsoft.com\u002Fazure\u002Ffoundry\u002Fagents\u002Fconcepts\u002Fcli-agent-development), and current [`microsoft-foundry\u002Ffoundry-samples`](https:\u002F\u002Fgithub.com\u002Fmicrosoft-foundry\u002Ffoundry-samples\u002Ftree\u002Fmain\u002Fsamples\u002Fpython\u002Fhosted-agents) as the sources of truth.\n\nDo not use older `Azure-Samples\u002Fazd-ai-starter-*` repositories as the reference implementation. Do not add deprecated standalone `agent.yaml` or `agent.manifest.yaml` files; the unified root `azure.yaml` replaces them.\n\nThis lab uses direct-code remote build through `codeConfiguration`. Do not add a Dockerfile, container registry workflow, or capability-host resources unless the lab deliberately changes to a container deployment or Standard Agent Setup.\n\n## Inspect Before Changing\n\nRead:\n\n- `src\u002Fagent\u002Fapp.py`\n- `src\u002Fagent\u002Frequirements.txt`\n- `azure.yaml`\n- `docs\u002Flab6-deploy-agent.md`\n- `.vscode\u002Ftasks.json`\n\nConfirm `azure.yaml` declares:\n\n- `azd` 1.27.1 or later\n- `azure.ai.agents` 1.0.0-beta.8 or later and `azure.ai.projects` 1.0.0-beta.4 or later under `requiredVersions`\n- an `azure.ai.project` service with the model deployment\n- `host: azure.ai.agent`\n- `uses` linking the agent to the project service\n- Python 3.13 under `codeConfiguration`\n- `app.py` as the entry point\n- Responses protocol `2.0.0`\n- `infra.provider: microsoft.foundry`\n\n## Preserve the Current Agent Pattern\n\n```python\nfrom agent_framework import Agent\nfrom agent_framework.foundry import FoundryChatClient\nfrom agent_framework_foundry_hosting import ResponsesHostServer\n\nagent = Agent(\n    client=FoundryChatClient(...),\n    instructions=SYSTEM_PROMPT,\n    default_options={\"store\": False},\n)\n\nResponsesHostServer(agent).run()\n```\n\nFoundry manages conversation continuity. Do not add local conversation-history persistence.\n\n## Prepare the Learner Environment\n\n1. Confirm `azd version` is 1.27.1 or later.\n2. Run `azd ext install microsoft.foundry`. This meta-extension installs the `azure.ai.*` providers; it does not replace their compatible `requiredVersions` entries in `azure.yaml`.\n3. Create and activate `.venv`.\n4. Install `src\u002Fagent\u002Frequirements.txt`.\n5. Run `python -m pip check`.\n6. Confirm Azure CLI and `azd` use the same tenant and subscription before cloud operations.\n7. Never print `.env`, access tokens, or connection strings.\n\n## Validate Locally\n\nUse the Foundry Toolkit F5 task when available. The local server is ready when:\n\n```text\nGET http:\u002F\u002F127.0.0.1:8088\u002Freadiness\n```\n\nreturns HTTP 200 with a healthy status.\n\n## Provision, Deploy, and Invoke\n\nThe first cloud operation runs both provider phases: `azd provision` creates the project and model deployment, then `azd deploy` publishes the hosted-agent version:\n\n```powershell\nazd up\n```\n\nThis operation can create billable Azure resources. State that before running it and require the learner's approval.\n\nFor later code-only updates to an environment that this project already provisioned:\n\n```powershell\nazd deploy\nazd ai agent show --output json\nazd ai agent invoke \"The delivery was late, but support kept me informed.\"\n```\n\nRemote deployment and invocation can incur Azure charges. State that before executing them. Do not run `azd up`, `azd deploy`, or `azd down` merely to test a theory.\n\n## Generate and Run Evaluations\n\nTreat evaluation as an optional post-deployment extension. Generation and evaluation are preview operations that make billable model calls, so state that before running them.\n\nUse `src\u002Fagent\u002Fevaluation-instructions.md` to generate a domain-specific synthetic suite. Keep `src\u002Fagent\u002Fevals\u002Fcaldova-golden.jsonl` as the curated regression set for regulated-routing expectations. Generated data broadens coverage; it does not replace human-reviewed golden cases.\n\n```powershell\nazd ai agent eval generate --agent caldova-consumer-sentiment-agent --gen-instruction-file src\u002Fagent\u002Fevaluation-instructions.md --eval-model gpt-5.4-mini --max-samples 15 --out-file eval.yaml\nazd ai agent eval run --config eval.yaml\n```\n\nInspect generated datasets, evaluators, and `eval.yaml` before running them. Do not run optimizer workflows automatically.\n\nWhen a schema criterion fails, inspect `sample.output_text` before changing the agent. Preview evaluators can mistakenly grade the serialized Responses transport envelope in `sample.output` or `sample.output_items`. If the assistant text is valid but the explanation cites wrapper fields such as annotations or output items, refine the evaluator to grade `sample.output_text`, upload it with `azd ai agent eval update --config eval.yaml --evaluator-only`, and rerun it. Do not present transport-envelope failures as agent regressions.\n\nIf wrapper-based failures persist or valid `sample.output_text` is marked not applicable, report the schema criterion separately as a preview evaluator limitation. Do not use the aggregate pass rate alone as the agent quality result.\n\n## Success Criteria\n\n- Agent dependencies resolve without conflicts.\n- The local readiness endpoint is healthy.\n- `azd up` provisions the project and creates an active hosted-agent version on first use.\n- `azd deploy` updates an active hosted-agent version without reprovisioning infrastructure.\n- Invocation returns governed JSON through the Responses protocol.\n- A learner can distinguish smoke invocation, generated evaluation coverage, and curated regression coverage.\n",{"name":5,"license":15,"description":7},"main","d8fef093482a4645e0e88611b9df43066a072f7f",".github\u002Fskills\u002Fill335-hosted-agent",11,"https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Ffrontiertour27-ILL335-building-deploying-ai-agents-with-microsoft-foundry","https:\u002F\u002Fgithub.com\u002Fmicrosoft","https:\u002F\u002Fopensource.microsoft.com","Open source projects and samples from Microsoft","2026-11-05 07:17:26",{"source":8,"skillName":5,"commands":55,"prompt":65,"zipAvailable":66},[56,59,62],{"tool":57,"command":58},"npx","npx skills add microsoft\u002Ffrontiertour27-ILL335-building-deploying-ai-agents-with-microsoft-foundry --skill ill335-hosted-agent",{"tool":60,"command":61},"bunx","bunx skills add microsoft\u002Ffrontiertour27-ILL335-building-deploying-ai-agents-with-microsoft-foundry --skill ill335-hosted-agent",{"tool":63,"command":64},"pnpm","pnpm dlx skills add microsoft\u002Ffrontiertour27-ILL335-building-deploying-ai-agents-with-microsoft-foundry --skill ill335-hosted-agent","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Ffrontiertour27-ILL335-building-deploying-ai-agents-with-microsoft-foundry\u002Fblob\u002Fmain\u002F.github\u002Fskills\u002Fill335-hosted-agent\u002FSKILL.md",true,[],[69,87,100,110],{"skillNo":70,"name":71,"nameZh":6,"description":72,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":73,"categories":74,"keywords":80,"publishedAt":40},"01M69BX3PZP4SV01RWPW6XCZMM","ill335-troubleshooting","Diagnoses ILL335 setup, authentication, dependency, Responses API, validation, and hosted-agent failures. Use when a learner reports an error, failed lab checkpoint, broken setup, dev container or Codespaces prebuild failure, deployment problem, rate limit, content filter, or unexpected model output.","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Ffrontiertour27-ILL335-building-deploying-ai-agents-with-microsoft-foundry\u002Fblob\u002Fmain\u002F.github\u002Fskills\u002Fill335-troubleshooting\u002FSKILL.md",[75,78,79],{"categoryNo":76,"categoryName":77,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000010","后端与接口",{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},{"categoryNo":27,"categoryName":28,"taxonomyType":29,"taxonomyTypeDesc":30},[81,82,83,84,38,85,86],"ill335 troubleshooting","dev","validation","authentication","codespaces","setup",{"skillNo":88,"name":89,"nameZh":6,"description":90,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":91,"categories":92,"keywords":95,"publishedAt":40},"01M69BX0P6SBE3HDJ26812XQY3","ill335-responses-api","Supports ILL335 model inference and sentiment-analysis labs that use the OpenAI Responses API with Microsoft Foundry. Use for Labs 3-5, responses.create, output_text, previous_response_id, structured JSON, model comparison, or Responses API errors.","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Ffrontiertour27-ILL335-building-deploying-ai-agents-with-microsoft-foundry\u002Fblob\u002Fmain\u002F.github\u002Fskills\u002Fill335-responses-api\u002FSKILL.md",[93,94],{"categoryNo":76,"categoryName":77,"taxonomyType":21,"taxonomyTypeDesc":22},{"categoryNo":27,"categoryName":28,"taxonomyType":29,"taxonomyTypeDesc":30},[96,97,98,99,34,33,9],"ill335 responses api","inference","openai responses api","create",{"skillNo":101,"name":102,"nameZh":6,"description":103,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":104,"categories":105,"keywords":106,"publishedAt":40},"01M69BWY0FW2SFXPSM09A7Q3Y9","ill335-lab-navigator","Guides learners through Frontier Tour ILL335, Building & Deploying AI Agents with Microsoft Foundry. Use when a learner asks where to start, what lab to complete next, how the labs connect, or how to resume the workshop.","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Ffrontiertour27-ILL335-building-deploying-ai-agents-with-microsoft-foundry\u002Fblob\u002Fmain\u002F.github\u002Fskills\u002Fill335-lab-navigator\u002FSKILL.md",[],[107,108,34,33,9,109],"ill335 lab navigator","connect","ai agents",{"skillNo":111,"name":112,"nameZh":6,"description":113,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":114,"categories":115,"keywords":116,"publishedAt":40},"01M69BWQZSR10T6TTKMJG2MNK7","ill335-devcontainer","Prepares and validates the ILL335 BYOD development container. Use for dev containers, GitHub Codespaces, Docker workspace setup, prebuilds, cached builds, rebuilds, container startup, virtual-environment conflicts, or workspace dependency installation failures.","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Ffrontiertour27-ILL335-building-deploying-ai-agents-with-microsoft-foundry\u002Fblob\u002Fmain\u002F.github\u002Fskills\u002Fill335-devcontainer\u002FSKILL.md",[],[117,118,82,119,85,86,120,121],"ill335 devcontainer","github","development","docker","containers"]