Azuresql Db Feedback

作者 microsoftc1e8167e4c1d無授權條款收錄於 2026年10月8日更新於 2026年10月8日

Reports a bug or files feedback about the azuresql-db-* agent skills themselves, or about the Azure SQL Database container (Private Preview). Use when the user says a skill or the container "did not work", hit an error, behaved unexpectedly, or is missing something; and when they say "report a bug", "file an issue", "open a GitHub issue", "request a feature", "give feedback", or "tell the team". Also use when you, the agent, had to deviate from an azuresql-db-* skill or work around a defect in one to finish a task: that is a bug worth reporting even if the task succeeded. Decides whether the problem belongs to the SKILL or to the CONTAINER, since they use different issue templates, then builds a complete prefilled GitHub issue from context you already have. Never submits anything without explicit confirmation from the user.

精選僅含說明Productivity & Workflow
AI 產生的概覽

將技能或容器問題整理成完整、已去識別化的 GitHub 議題草稿,供使用者確認後提交。

功能
此技能引導代理判斷問題屬於 azuresql-db-* 技能還是 Azure SQL Database 容器,接著蒐集相關事實,例如技能名稱、代理執行環境、映像標籤、主機作業系統與重現指令。它會將 SA 密碼、登錄檔憑證與連線字串等機密資訊去識別化,起草預先填好的 GitHub 議題,並在提交前完整顯示給使用者。提交會先嘗試 gh CLI,其次使用預填的 github.com 網址,且未經使用者明確確認不會提交任何內容。
適用情境
當使用者回報某個 azuresql-db-* 技能或 Azure SQL Database 容器無法運作、發生錯誤、行為異常或缺少功能時,或當使用者要求回報缺陷、提交議題或提供意見回饋時使用。當代理為了完成任務而必須偏離技能說明或繞過某個缺陷時,也適用。
執行需求
僅為說明文件,不附帶指令碼。第一層提交方式需要 gh CLI 且 gh auth status 成功;第二層預填 github.com 網址方式不需要任何憑證。建構任何網址前需讀取 references/issue-fields.md,可選用 Microsoft Learn MCP 伺服器取得最新文件。

Report a problem with the skills, or with the container

The user should never have to assemble a bug report by hand. You are already holding what a good one needs: which skill you were following, what it told you to do, what you actually had to do, the image tag, the host, the runtime, the failing command, and the error. Your job is to turn that into a complete report and hand it to the user to submit.

Repository: microsoft/azure-sql-database-container (both the container and the skills live here).

The claims in this skill are about issue forms in microsoft/azure-sql-database-container, not about the engine, so no engine measurement applies to them. Verified on 2026-09-05: aka.ms/sql-agent-skills-feedback still answers 301 to https://github.com/microsoft/azure-sql-database-container/issues/new?template=skill_feedback.yml with the query string dropped, which is exactly why a prefilled report has to use the full github.com URL. The field ids, labels and dropdown values are read from the issue form definitions in this repository and mirrored in references/issue-fields.md [blocked], which you open before you build any URL.

The two rules that matter most

  1. Never create an issue without explicit confirmation. Show the user the full title and body first and wait for a clear yes. You may offer; you may never submit unasked. This applies to every path below, including the gh CLI.
  2. Redact before you send. Everything submitted is public. See Redaction.

Step 1: skill problem, or container problem?

Get this right first: they use different issue templates.

Ask yourself what actually failed.

The problemIt belongs toTemplate
A skill told the agent to do the wrong thing, or failed to say something it needed to saythe skillskill_feedback.yml
The wrong skill loaded, or no skill loaded when one should havethe skillskill_feedback.yml
A skill would not install, or the agent never picked it upthe skillskill_feedback.yml
You had to deviate from the skill or work around it to make the task succeedthe skillskill_feedback.yml
The skill said the right thing, the agent did the right thing, and the container still failed (will not start, query errors, engine behaves unexpectedly)the containerbug_report.yml
The container is missing a capability the user wantsthe containerfeature_request.yml

The test that settles it: if the instructions were correct and the engine still broke, it is a container bug. If the instructions were wrong, incomplete, or ignored, it is a skill bug. When genuinely torn, ask the user. Do not guess, and do not file both.

If the user is only reporting that something worked well, do not open an issue. Point them at Discussions and stop.

Step 2: gather the facts

Only run a command if you do not already know the answer.

For a skill problem (the important ones, and the ones we are otherwise blind to):

  • Which skill, by name (azuresql-db-rag, etc). If none loaded, say so: that is itself the bug.
  • Which agent you are (Claude Code, GitHub Copilot, Codex, Cursor). Skills behave differently across harnesses and we cannot see which one the user ran.
  • The instruction that was wrong or missing, quoted from the skill, and what actually worked instead. This is the single most valuable field in the whole report. Without it we cannot fix the skill.
  • The prompt the user gave you.

For a container problem: the image tag, the host OS (mapped to the exact dropdown value), the container runtime and version, what happened, and the commands that reproduce it. docker logs sqldb if the container is what broke.

Step 3: redaction (do this before you build anything)

Strip all of the following from every field, including logs and repro steps:

  • The SA password (MSSQL_SA_PASSWORD, -P <password>, the Password= field of a connection string). Replace with ***.
  • Registry credentials (the docker login username and password).
  • Full connection strings. Keep the shape, drop the secret: Server=localhost,1433;Database=appdb;User Id=sa;Password=***;TrustServerCertificate=true.
  • Any customer or application data, table contents, embeddings, or paths that identify the user or their employer.
  • Access tokens of any kind.

The repository is public and issues are world readable. A report that leaks the user's password is worse than no report at all.

Step 4: draft and show

Write the report, then show it to the user in full and ask whether to file it. Say which fields you filled and which you left blank. If you could not determine a value confidently, leave it blank rather than guess: a wrong skill name or host OS sends triage down the wrong path.

Step 5: submit

Try these in order, stop at the first that works, and confirm with the user first either way.

Tier 1: the gh CLI, if gh auth status succeeds. The issue is authored by the user, so they get replies.

bash
gh issue create --repo microsoft/azure-sql-database-container \  --title "[Skill]: <one-line summary>" \  --label skills --label needs-triage --label User-filled --label via-skill \  --body-file <path-to-body>

Write the body to a file rather than passing it inline, so quoting and newlines survive.

Tier 2: a prefilled URL. No credentials, no setup, and the user sees exactly what will be submitted before anything happens. Use this whenever gh is not available; it always works.

Build the URL from the issue form's field ids and print it for the user to open. The field ids, the verbatim dropdown values, and worked examples for all three templates are in references/issue-fields.md [blocked]. Read that file before constructing a URL.

Validation rules

  • The user saw the exact title, body and labels before anything was submitted, and said yes.
  • The report went to the right template: a skill problem on the skills form, a container problem on the container form.
  • No SA password, registry credential or access token appears in any field. Re-read the body once more before submitting.
  • Every dropdown value is verbatim from the field reference, which you open before building any URL, or the field was left out.
  • A prefilled report uses a full github.com URL, never an aka.ms link, because those drop the query string.
  • Tier 1 succeeded when gh issue create printed the new issue URL; tier 2 succeeded when the user opened the URL and the form came up with the fields already filled in. If neither happened, the report was not filed: say so rather than implying it was.

Do not

  • Do not create, comment on, or close an issue without explicit user confirmation.
  • Do not include the SA password, registry credentials, or access tokens in any field. Ever.
  • Do not file a skill problem on the container template, or the reverse. They are triaged by different people.
  • Do not guess a dropdown value. Leave the field out if you are unsure; an omitted dropdown renders unselected, but a wrong one misroutes triage.
  • Do not use an aka.ms short link to carry a prefilled report. Those links drop query strings, so every field you filled in is silently lost. Use the full github.com URL. (The aka.ms links are for humans opening an empty form: skills feedback, container bug, container feature.)
  • Do not open an issue for something already on the Known limitations page, or already in open issues. Point the user there instead.
  • Do not nag. Offer once, take no for an answer, and move on.

References

  • references/issue-fields.md [blocked]: the exact field ids for all three templates, the verbatim dropdown values, URL construction and its length limit, and worked examples. Read this before building a prefilled URL.

Staying current

Authoritative, version-pinned references for the tools this skill uses (read the one you need):

If the Microsoft Learn MCP server is configured, use mcp__microsoft-learn__microsoft_docs_search or mcp__microsoft-learn__microsoft_docs_fetch to fetch the current version of any of these on demand. It is optional; when it is unavailable, the references above are authoritative.

來源與署名

來源:microsoft/azure-sql-database-container位於skills/azuresql-db-feedback提交c1e8167

授權條款: 無授權條款

內容歸原作者所有。SourceWeft 從公開儲存庫中收錄這些內容。

檢舉或申請下架

更多來自 microsoft/azure-sql-database-container 的技能

Azuresql Db Sidecar

microsoft

精選

將 Azure SQL 資料庫容器以 sidecar 服務形式加入 Docker Compose 或 Dev Container。

DevOps & Cloud2026年10月8日

Azuresql Db Rag

microsoft

精選

Builds local vector search, RAG, embeddings, and semantic search on the Azure SQL Database container using the native VECTOR type and VECTOR_DISTANCE. Use when you need to store embeddings, do similarity search, top-k nearest neighbor, cosine distance, retrieval-augmented generation, "find similar documents", chatbot memory, or semantic lookup against a local SQL database. Use this instead of pgvector, FAISS, Chroma, Pinecone, or a separate vector store when the data already lives in (or can live in) Azure SQL. Covers the VECTOR(n) column type, inserting embeddings with CAST(CAST(? AS NVARCHAR(MAX)) AS VECTOR(n)) where the dimension is a literal, a pluggable embed() so only the endpoint changes for cloud, and a working CREATE VECTOR INDEX with the two errors that block it. Provisions appdb on master first so every script runs on a fresh container.

待分類2026年10月8日

Azuresql Db Import

microsoft

精選

使用 SqlPackage 將 .bacpac 或 .dacpac 匯入本機 Azure SQL Database 容器。

DevOps & Cloud2026年10月8日

Azuresql Db From Sql Server

microsoft

精選

將本機 SQL Server Docker 環境遷移至 Azure SQL Database 容器,以實現與 Azure 一致的本機開發。

DevOps & Cloud2026年10月8日

Azuresql Db Faq

microsoft

精選

解答 Azure SQL 資料庫容器(個人預覽版)支援哪些功能,以及為何與雲端服務不同。

Learning & Education2026年10月8日

Azuresql Db Ci

microsoft

精選

Runs integration tests against the Azure SQL Database container (Private Preview, local engine) in CI. Use when setting up GitHub Actions, Azure Pipelines, or GitLab CI to test against Azure SQL DB; when adding a database service container to a CI workflow; when tests need a real Azure SQL engine in the pipeline; or when you see "service container", "health-cmd", "ACR_USERNAME/ACR_PASSWORD", "MSSQL_SA_PASSWORD secret", or "integration test database". Also use when a workflow was about to pull the SQL Server image mcr.microsoft.com/mssql/server, in which case stop and use the Azure SQL Database engine image instead. Covers pulling from the private ACR with credentials, the service health check that runs sqlcmd inside the container so the runner needs no client tools, provisioning appdb before tests, and pointing the test connection string at the user database not master.

待分類2026年10月8日