[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M69AGJC50G0YX7VQ52SQ0AD0":3,"skill-same:01M69AGJC50G0YX7VQ52SQ0AD0":113},{"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":28,"publishedAt":35,"linkOnly":36,"markdownContent":37,"markdownBody":38,"frontMatter":39,"refName":40,"commitSha":41,"skillDirectory":42,"forksCount":43,"repositoryHtmlUrl":44,"repositoryHomepageUrl":6,"ownerHtmlUrl":45,"ownerBlogUrl":46,"ownerDescription":47,"updatedAt":48,"install":49,"resources":62},"01M69AGJC50G0YX7VQ52SQ0AD0","sql-bacpac-to-azure-sql-db-migration",null,"Migrate one or all user databases from SQL Server to Azure SQL Database with an offline BACPAC workflow. Use for BACPAC, SqlPackage export\u002Fimport, bulk database migration, or offline Azure SQL Database migration requests.","microsoft\u002Fsqlcon-eu-2026-demos","microsoft","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F6154722?v=4","01M698VYDXZ6JW5WH3CRV650R8","Microsoft","01M69AFC8P5962CJBG1RDAKA7Q",3,"MIT","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fsqlcon-eu-2026-demos\u002Fblob\u002Fmain\u002Fcorenote\u002Fmigration\u002Fskills\u002Fsql-bacpac-to-azure-sql-db-migration\u002FSKILL.md",[18,23],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000010","后端与接口","SCENARIO","使用场景",{"categoryNo":24,"categoryName":25,"taxonomyType":26,"taxonomyTypeDesc":27},"01JBSEEDCAT000000000000033","数据分析师","PROFESSION","适用职业",[29,30,31,32,33,34],"azure sql","sql","sql bacpac to azure sql db migration","workflow","azure sql database","sql server","2026-11-10 10:12:30",false,"---\nname: sql-bacpac-to-azure-sql-db-migration\ndescription: \"Migrate one or all user databases from SQL Server to Azure SQL Database with an offline BACPAC workflow. Use for BACPAC, SqlPackage export\u002Fimport, bulk database migration, or offline Azure SQL Database migration requests.\"\n---\n\n# SQL Server to Azure SQL Database using BACPAC\n\n## When to Use\n\n- Migrate one or more SQL Server user databases to Azure SQL Database using BACPAC export and import.\n- Perform an offline Azure SQL Database migration with `SqlPackage`.\n\nHelp a user migrate one or all user databases from **SQL Server** to **Azure SQL\nDatabase** using an offline **BACPAC** workflow. A BACPAC contains the database schema and data. It\ndoes not preserve instance-level objects, server configuration, SQL Agent jobs,\nor a transactionally consistent online migration.\n\nUse this skill when the user asks to migrate a SQL Server database to Azure SQL\nDatabase with BACPAC, `SqlPackage`, import\u002Fexport, or an offline database move.\n\nRead [references\u002Fcommand-execution-setup.md](references\u002Fcommand-execution-setup.md) and\n[references\u002Fcommand-execution-export-import.md](references\u002Fcommand-execution-export-import.md)\nbefore generating or running commands.\nResolve the absolute installed directory containing those reference files and\npass it to the command blocks as `$skillReferenceRoot`. Do not derive that path\nfrom the current directory or `$PSScriptRoot`; the orchestration runs inline,\nnot from a script file.\n\nThe `Invoke-SqlPackageWithProgress` implementation in the checked-in setup\nreference is the only permitted process boundary for `SqlPackage` export and\nimport. Do not replace it with `Start-Process`, a shell command string, an\nalternate script, or an executor-built wrapper. It uses\n`ProcessStartInfo.ArgumentList` so every complete `\u002FName:\u003Cvalue>` element remains\none operating-system process argument and rejects empty or detached\nconnection-string and BACPAC-path fragments before launch.\nUse only the manifest, checkpoint, and per-database folder layout defined by the\nreference blocks; do not introduce aggregate `exports`, `manifest`, `evidence`,\nor `checkpoint` folders.\n\nValidate modern Go `sqlcmd` capabilities from the combined output of every\nsuccessful supported help mode (`--help` and `-?`). A required flag may appear\nin either output. Report the successful help modes and exact missing flags before\noffering installation; do not treat a reduced `--help` view as proof that an\ninstalled client lacks flags exposed by `-?`. Quote the compatibility argument as\n`'-?'`, and name wrapper parameters `$ArgumentList`, never `$Args`, so PowerShell's\nautomatic `$args` variable cannot consume the intended arguments. After an\napproved `winget install sqlcmd`, rerun capability validation regardless of the\nwinget exit code; an already-installed\u002Fno-upgrade result is successful when the\nrefreshed client passes validation.\n\n\n## Operating contract\n\n- Collect all missing nonsecret migration inputs in one grouped request whenever\n  possible, except database selection must follow authoritative source discovery\n  when the user has not already supplied an exact database name or `All`. Reuse\n  values already supplied by the user or discovered safely.\n- Treat this as an **offline migration**. State that downtime will occur, but do\n  not request a separate downtime acknowledgment. Include that acknowledgment in\n  the single target-database creation approval. Do not ask when downtime should\n  start, how long it may last, or what duration is acceptable.\n- Require the user to select one online user database or all selectable online\n  user databases. Never interpret `all` as including system databases.\n- Do not run source migration readiness, compatibility, feature, dependency,\n  performance, or sizing assessments. Even if the database is online and in read-write mode, allow the export to continue. However, inform the user that they must ensure there is no DML  and DDL activity on the database during the export.Consume prior skill outputs when supplied;\n  otherwise use the source and target decisions provided by the user.\n- For target details, require the existing Azure SQL logical server name and an\n  explicit target configuration after export: service type (`GeneralPurpose`,\n  `BusinessCritical`, or `Hyperscale`), service objective, and vCore count. Apply\n  the catalog's documented maximum size automatically. Before export, collect only the existing logical server\n  name. After every export is attempted and at least one succeeds, track each\n  user input separately. Never infer a representative objective or vCore count\n  from the service type. Validate the resolved configuration with\n  `Resolve-ApprovedTargetSkuSelection`; do not accept an arbitrary objective or\n  size outside the catalog. Do not ask for subscription, resource group, region, or target\n  authentication before export.\n- Never connect to either the source or target with SQL login\u002FSQL authentication.\n  Do not request, accept, retrieve, or use SQL passwords or SQL credentials. Use\n  only Windows Integrated authentication or Microsoft Entra authentication\n  supported by the workflow. Never read SQL credentials from Key Vault or\n  Credential Manager, or expose, print, commit, or persist them.\n- Run each `SqlPackage` operation with bounded status polling. Emit a progress\n  update every 30 seconds with the database, batch position, operation, elapsed time, process\n  state, diagnostics\u002FBACPAC file growth, and latest sanitized diagnostic status\n  so the user can distinguish movement from a potentially stalled operation.\n  Write each static heartbeat to both the live terminal and a sanitized\n  operation-status file beside the diagnostics, and flush host output after each\n  line. If the command runner buffers terminal output until completion, tail the\n  status file and relay its latest line at least every 30 seconds; do not rely on\n  the transient `Write-Progress` display as the only visible progress channel.\n  Print an immediate `Started` message before the first poll and tell the user\n  that live status appears in the terminal. After five minutes without diagnostic\n  or BACPAC growth, report `Possibly stalled` while making clear that the process\n  is still active. Do not invent a percentage when `SqlPackage` does not provide one.\n- Pass complete `\u002FSourceConnectionString:\u003Cvalue>`,\n  `\u002FTargetConnectionString:\u003Cvalue>`, `\u002FSourceFile:\u003Cvalue>`, and\n  `\u002FTargetFile:\u003Cvalue>` forms as individual string-array elements to the\n  checked-in progress wrapper. Never split a switch name from its value or\n  flatten the argument array into a command string.\n- Do not execute destructive operations against the source database.\n- After all exports and target authentication, run one injection-safe, read-only\n  target `sys.databases` preflight through validated Go `sqlcmd` for the selected\n  names. Mark databases observed\n  before import as skipped. Never infer this condition from localized SqlPackage\n  diagnostics.\n  - Do not perform any write or edit or perform any DDL opertion in the source databases\n- Surface errors and record the failed phase. For an `All` export or import, skip\n  the failed database and continue the remaining databases. Do not report success\n  based on a command that was not confirmed successful.\n- Use Windows PowerShell syntax when running local commands on Windows.\n\n## Assessment boundary\n\nThis skill assumes a separate assessment or user decision has already selected\nAzure SQL Database, BACPAC, and any required source remediation.\nDo not inspect the source for migration blockers or reevaluate the selected target\nand method. If a prerequisite is missing, route to the appropriate assessment or\nplanning skill rather than adding assessment prompts to this workflow.\n\nAn export error may identify an unsupported object or feature. Record the error\nand skip that database when processing an `All` selection; do not broaden the\nmigration run into a source assessment.\n\n\n## Workflow\n\n```text\nPhase 0: Validate tooling, then collect source connection and safety inputs\nPhase 1: Discover databases, resolve source scope, and prepare the complete execution manifest\nPhase 2: Prepare local folders without an approval prompt\nPhase 3: Export each selected database to its BACPAC subfolder\nPhase 4: Select target SKU and size, authenticate, approve, then import with SqlPackage\nPhase 5: Hand every successfully imported database to post-migration validation\n```\n\n## Phase 0 — Collect inputs and safety gates\n\nTreat modern Go-based `sqlcmd` as a mandatory prerequisite for the BACPAC\nworkflow. Before discovery, folder creation, or export, inspect every `sqlcmd`\nexecutable on `PATH`; do not assume the first result is modern because ODBC and\nGo-based versions can coexist. Select only a candidate whose `--version` command\nsucceeds and whose help exposes `-G`, `-U`, and `--authentication-method`. If no\ncandidate passes, ask once for approval to install modern `sqlcmd` with the official\n`winget install sqlcmd` command. Do not install silently. If approved, install it,\nrefresh the current process `PATH`, and repeat the checks. If installation is\ndeclined or validation still fails, stop before starting the migration and provide\nthe official manual-download link. Cache the resolved full path for Phase 4.\n\nDo not ask for target service type, objective, vCore, or maximum size in Phase 0.\nAfter Phase 3 has attempted every selected export and at least one succeeds, load\nthe checked-in catalog\n`references\u002Fazure-sql-db-vcore-sku-limits.json` with\n`Get-ApprovedTargetSkuCatalog` from `references\u002Ftarget-sku-validation.ps1`.\nThe catalog is derived from Microsoft Learn's *Single database vCore resource\nlimits* page, source-dated March 9, 2026 and updated April 10, 2026. Use the\nstaged input tracker and make collection interactive:\n\n1. Present exactly `General Purpose`, `Business Critical`, and `Hyperscale` as\n  service-type selections.\n2. After that selection, filter the approved catalog and present only its exact\n  provisioned\u002Fserverless hardware objective labels for the chosen type. Do not\n  present a DTU `Standard`\u002F`Premium` tier as a vCore objective.\n3. After the objective selection, load its approved vCore values and require an\n  explicit integer selection\u002Finput.\n4. Then display the documented maximum data size for that exact vCore and apply\n  it automatically without a separate confirmation prompt. Catalog entries must provide a\n  `DocumentedMaxDataSizeGBByVCore` map. It records the resource-limit ceiling,\n  not every configurable storage increment.\n\nKeep separate status\u002Fvalue\u002Foptions entries for `ServiceType`,\n`ServiceObjective`, `VCore`, and `MaximumSizeGB`. Changing service type clears\nthe objective, vCore, and size; changing objective clears vCore and size. Never\nselect the first option or infer `1024 GB` for Hyperscale. The approved catalog\nmust map the friendly objective and vCore to an exact\n`DatabaseServiceObjectiveTemplate` used to derive the SqlPackage objective\n(for example `HS_Gen5_{0}` plus vCore `4` becomes `HS_Gen5_4`). Validate the\ncomplete selection through `Resolve-ApprovedTargetSkuSelection`. Do not use\nSqlPackage import to discover that a SKU is unavailable for the target server,\nregion, subscription, quota, or capacity.\n\nThe checked-in stable SKU list is:\n\n| Service type | Compute\u002Fhardware objective | Published vCore choices | Published max data size |\n| --- | --- | --- | --- |\n| General Purpose | Provisioned standard-series (Gen5) | 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 24, 32, 40, 80, 128 | 1–4 TB, depending on vCore |\n| General Purpose | Serverless standard-series (Gen5) | 1, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 24, 32, 40, 80 max vCores | 512 GB–4 TB, depending on max vCore |\n| General Purpose | Provisioned DC-series | 2, 4, 6, 8 | 1–3 TB, depending on vCore |\n| Business Critical | Provisioned standard-series (Gen5) | 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 24, 32, 40, 80, 128 | 1–4 TB, depending on vCore |\n| Business Critical | Provisioned DC-series | 2, 4, 6, 8 | 768 GB |\n| Hyperscale | Provisioned standard-series (Gen5) | 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 24, 32, 40, 80 | 128 TB |\n| Hyperscale | Serverless standard-series (Gen5) | 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 24, 32, 40, 80 max vCores | 128 TB |\n| Hyperscale | Provisioned DC-series | 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 32, 40 | 128 TB |\n| Hyperscale | Provisioned premium-series | 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 24, 32, 40, 64, 80, 128 | 128 TB |\n| Hyperscale | Provisioned premium-series memory optimized | 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 24, 32, 40, 64, 80 | 128 TB |\n\nExclude Fsv2-series because it is unavailable for new creation and scheduled for\nretirement on October 1, 2026. Exclude preview choices, including General Purpose\nand Business Critical DC-series 10–40 vCores and Hyperscale premium-series 160\nand 192 vCores. Treat every availability note as a guardrail: the catalog does\nnot prove regional availability, quota, subscription eligibility, or capacity.\n\nProduce a concise migration plan before execution:\n\n| Item | Value |\n| --- | --- |\n| Source Server | `\u003Cexisting SQL Server instance name>` |\n| Target server | `\u003Cexisting Azure SQL logical server name>` |\n| Target configuration | `Selected after export` |\n| Method | Local BACPAC export\u002Fimport with `SqlPackage` |\n| Downtime | Offline; acknowledged in the target-database creation approval |\n| Overwrite behavior | Create new database; never overwrite by default |\n| Local root | `\u003Cuser-provided path or default path>` |\n\nCollect the missing source authentication, existing target logical server, and\nexport root in one grouped request. Source authentication is never inferred from\nthe host or silently defaulted: require an explicit selection of `Windows\nIntegrated` or `Microsoft Entra Interactive MFA`. For the export root, show\n`$env:USERPROFILE\\SqlMigration\\Bacpac` as the recommended default and require the\nuser either to accept it or provide another absolute path; never append a\ntimestamp or create a run directory before this choice is resolved. Reuse a value\nonly when the user supplied it explicitly in the current request or earlier in\nthe conversation.\n\nBefore discovery or any filesystem write, create an in-memory input-resolution\nrecord containing `SourceServer`, `SourceAuthentication`, `TargetServer`, and\n`ExportRoot`. If any field is unresolved, stop and ask the grouped question. Do\nnot run discovery, create directories or checkpoints, or invoke `SqlPackage`\nuntil all four fields are resolved. Echo the resolved nonsecret values to the\nuser before proceeding.\n\nIf the user's original request already names one exact source database or says\n`All`, retain that value as the requested selection and do not ask for it again\nbefore discovery. Otherwise, defer the database-selection field until Phase 1 so\nit can contain the discovered database names. Input collection is not execution\napproval. Do not request approval in Phases 0-3, and do not collect a downtime\nstart time, window, maximum duration, RPO, or RTO.\n\n## Phase 1 — Discover databases and build the manifest\n\nConnect to the source database using the following authentication methods:\na. Windows Authentication\nb. Entra Authentication using Interactive MFA sign-in\n\nAfter source authentication is established, always discover all online user\ndatabases from the source instance before presenting a database-selection form.\nExclude `master`, `model`, `msdb`, `tempdb`, snapshots, and offline databases.\nThis is scope discovery, not migration assessment.\n\nResolve the database scope as follows:\n\n1. If the user's original request or prior answer already supplied one exact\n   database name, match it against the discovered selectable list. When it\n   matches, reuse it without asking again. When it does not match, show the\n   discovered list and ask for a corrected selection.\n2. If the user already supplied `All`, reuse it without asking again; `All`\n   means every database in the discovered selectable list.\n3. Otherwise, display a single-choice form whose options are `All` followed by\n   every exact discovered database name. Never display `All` as the only preset\n   when one or more selectable databases were discovered, and do not require the\n   user to type a discovered database name as free text.\n\nIf no selectable online user databases are discovered, stop with a clear error\ninstead of showing an empty or `All`-only selection form.\n\n\nRequire the source authentication selection made in the Phase 0 grouped request.\nIf Microsoft Entra authentication is selected, use the supported `SqlPackage`\nauthentication options for the installed version.\nReject SQL login authentication and do not request or use a password from any\nsource, including Key Vault, Credential Manager, masked controls, or `sqlcmd`.\n\nResolve the export root from the grouped input request. The recommended default is\n`$env:USERPROFILE\\SqlMigration\\Bacpac`, but it is not selected until the user\naccepts it. Use the accepted path exactly as the export root; do not silently add\na timestamped child folder. Build the complete in-memory manifest and database\nmappings without creating folders, databases, or BACPAC files.\n\n\n## Phase 2 — Prepare without prompting\n\nDisplay the final source scope, target server name, database mappings, local root,\nand overwrite behavior. State clearly that the workflow is offline and will cause\ndowntime.\nDo not request execution confirmation here. Prepare local folders and run every\nselected export from the displayed manifest. Do not prompt between databases,\nfor authentication retry routing, after failures, or between export and import.\nAn authentication or unsupported-object failure is not remediated inline: save\nthe database as awaiting manual remediation, give concrete external next steps,\nand continue only with unaffected databases when `All` was selected.\n\nStop with a clear error when required nonsecret input is missing or inconsistent.\nChanging application connection strings remains a separate cutover action.\n\nThis skill does not discover, create, or configure Azure resources. The target\nlogical server must already exist. Do not open a target SQL data-plane connection,\nrequest target credentials, or test target database authentication before every\nselected export has been attempted. The import phase creates one database for each\nsuccessfully exported source database after confirming that the name is not in use.\n\n### Prepare local folders\n\nFor every selected database, use this exact layout:\n\n```text\n\u003Cexport-root>\\\n  \u003Cdatabase-name>\\\n  \u003Cdatabase-name>.bacpac\n  attempts\\\n    export\\\u003CUTC-GUID-attempt-id>\\\n      diagnostics.log\n      console.log\n      status.log\n    import\\\u003CUTC-GUID-attempt-id>\\\n      diagnostics.log\n      console.log\n      status.log\n```\n\nThe database name must be valid as a Windows folder name. If it contains an\ninvalid path character, is a reserved Windows name, would resolve outside the\nexport root, or maps to an existing symbolic link or junction, exclude that\ndatabase and display a warning that names the database and the reason it was\nskipped. Do not silently sanitize it because the subfolder name controls the\ntarget database name. If no eligible databases remain, stop before creating\nfolders.\n\nPersist the user's original database selection as `SelectionMode`, with the\nvalue `All` or `Explicit`; never infer this mode from the number of manifest\nentries after discovery or path filtering. The migration manifest contains\n`SourceDatabase`, `FolderPath`, `BacpacPath`,\n`TargetDatabase`, `ExportStatus`, `ExportFailureReason`, `ExportLastUpdatedUtc`,\n`ImportStatus`, `ImportFailureReason`, `ImportLastUpdatedUtc`,\n`ImportStartedUtc`, `ImportCompletedUtc`, `ImportDuration`,\n`ImportLastActivityUtc`, `TargetStateAfterImport`, `ResumeState`,\n`FailureCategory`, `ManualNextAction`, and\n`ValidationReportStatus`. Set `TargetDatabase` to the subfolder name. Use this\nsame manifest through export, import, and the validation handoff; do not rediscover\nor broaden the selection later.\n\nUse section 2 of [references\u002Fcommand-execution-setup.md](references\u002Fcommand-execution-setup.md) after the\nmanifest is displayed; no approval is required to create local folders.\n\n## Phase 3 — Export with SqlPackage\n\nUse the latest supported `SqlPackage` installed on the operator machine.\nValidate the executable and version before running the export.\n\nUse the export loop in section 3 of\n[references\u002Fcommand-execution-export-import.md](references\u002Fcommand-execution-export-import.md) and obtain connection\ndetails from an approved secure provider.\n\nTreat the authenticated source connection as an immutable batch-level base. For\nevery manifest row, create a fresh connection-string builder, remove both\n`Database` and `Initial Catalog` aliases, and set `Initial Catalog` to that row's\n`SourceDatabase` unconditionally. A catalog supplied by the base connection,\nincluding `master`, must never determine which database is exported.\n\nSet the manifest status to `InProgress` before starting `SqlPackage`. Poll the\nprocess every 30 seconds, update `ExportLastUpdatedUtc`, and display a console\nheartbeat without exposing arguments or credentials. Include whether diagnostic\nor BACPAC bytes increased since the previous poll and the latest sanitized\ndiagnostic status line. File growth is evidence of movement, not a completion\npercentage; if neither file grows, state that the process is still running rather\nthan claiming it is stuck.\n\nWhen the process exits, display `Succeeded` or `Failed`\nimmediately. On failure, include the database, exit code, sanitized reason, and\ndiagnostics path in the user-facing update.\n\nFor Microsoft Entra authentication, use the supported `SqlPackage`\nauthentication options for the installed version. Do not put an access token in\nthe command line.\n\nFor Windows integrated authentication, place `Integrated Security=True` (or\n`Trusted_Connection=True`) inside `\u002FSourceConnectionString`. Never generate\n`\u002FSourceIntegratedSecurity`; it is not a supported `SqlPackage` 170.4 argument.\nValidate the complete argument template before the first database export so an\nunsupported switch cannot fail multiple databases. If an already-running batch\nreports an unrecognized argument, stop that batch, report the configuration\nfailure immediately, save the checkpoint, and provide the installed SqlPackage\nversion plus the unsupported argument as manual correction guidance. Do not edit\nor rerun the command inline. Resume only failed\u002Fpending databases after the user\nconfirms the external configuration correction; do not re-export successful ones.\nTreat configuration correction as manual remediation; the confirmation phrase is\n`source remediation complete` because it unblocks the export stage, even though\nno source database change is authorized.\n\nExport gates:\n\n- Immediately before each export, check the exact manifest `BacpacPath` with a\n  literal-path file test. If a file already exists at that path, do not start\n  SqlPackage and never overwrite, rename, delete, or reuse the file. Mark that\n  export `Failed` and `AwaitingManualRemediation`, report the full path and\n  `A BACPAC file with the same name already exists`, save the checkpoint, and\n  instruct the user to preserve\u002Fmove the existing evidence or choose a new export\n  root outside this workflow. Resume only after `source remediation complete`\n  and after confirming that the intended path no longer exists. For `All`,\n  continue only with databases whose BACPAC destination does not already exist.\n- Confirm the export exit code is successful.\n- Confirm the export produced the expected BACPAC file. Artifact validation is\n  deferred until immediately before import.\n- Confirm the source was not modified by the export workflow.\n- Record the export timestamp and source database state.\n\nIf export fails, preserve the diagnostics and record the database and failure\nreason. Never silently omit a failed database or failed object from the report.\n\nIf diagnostics identify an authentication failure, do not refresh credentials,\nretry, prompt for another authentication method, or switch to an Entra\u002FSQL\u002Fmanaged\nidentity fallback in the current execution. Mark the export `Failed`, set\n`ResumeState = AwaitingManualRemediation`, save a sanitized checkpoint, and show:\n\n1. the selected authentication method and sanitized SqlPackage error;\n2. the diagnostics and checkpoint locations;\n3. instructions to verify the same identity can connect to and read the source\n  database, correct its login\u002Fuser permissions, firewall\u002Fnetwork route, MFA or\n  approved secure-store entry outside this workflow as applicable; and\n4. instructions to return with `authentication remediation complete` to resume\n  the failed database, or restart with the same export root.\n\nNever request, print, or persist a password or token while reporting the failure.\nDo not retry until the user confirms that external remediation is complete; then\nrevalidate the connection and retry only the failed\u002Fpending database. A different\nauthentication method is a restarted\u002Frevised run, not inline remediation.\n\nIf diagnostics identify unsupported source schema, code, or objects, do not alter\nDDL, drop\u002Fdisable an object, restore or clone another database, export from a\nsubstitute database, or invoke an automated remediation. Record the exact\nunsupported item when diagnostics provide it, set the same awaiting-remediation\nstate, and direct the user to review the diagnostics, assess and approve the\nrequired source change outside this skill, back up the source, apply and test that\nchange manually, and return with `source remediation complete`. Do not invent a\ngeneric schema change when SqlPackage does not identify one.\n\nFor an `All` selection, mark that database `Failed`, skip it, and continue exporting\nthe remaining databases. After every selected export has been attempted, proceed\nwith only manifest items whose `ExportStatus` is `Succeeded`. If no exports\nsucceeded, stop before target connectivity. For a single-database selection, stop\nwhen its export fails. After the export loop, display a summary of succeeded and\nfailed databases as a table with `DatabaseName`, `Status` (`Success`, `Fail`, or\n`InProgress`), and human-readable `BacpacSize`. Show `Not created` when no BACPAC\nfile exists. Include each failed database's sanitized reason and diagnostics path\nbefore starting target connectivity.\n\n## Phase 4 — Initiate target connectivity and import\n\nAfter every selected export has reached `Succeeded` or `Failed` and at least one\nBACPAC succeeded, reuse the modern Go-based `sqlcmd` executable validated in\nPhase 0 and invoke it by its resolved full path. If that exact executable is no\nlonger available, stop before target connectivity and report that the prerequisite\nchanged during the run; do not substitute an unvalidated executable.\n\nBefore target authentication or connectivity, select and freeze the target SKU\nand size using these steps:\n\n1. Load the checked-in catalog with `Get-ApprovedTargetSkuCatalog`.\n2. Display `General Purpose`, `Business Critical`, and `Hyperscale`; require one\n  service type.\n3. Filter and display only the compute\u002Fhardware objectives for that service type;\n  require one objective.\n4. Display only the published vCore values for that objective; require one integer.\n5. Apply the Microsoft Learn documented maximum data size for that vCore\n  automatically; do not present a separate Yes\u002FNo acceptance flow.\n6. Call `Complete-TargetConfiguration`, display the resolved service type,\n  objective, vCore, maximum size, `DatabaseEdition`, and\n  `DatabaseServiceObjective`, then persist them in the sanitized checkpoint.\n\nIf a Phase 4 checkpoint already contains a target configuration, revalidate and\nreuse it rather than asking again. If the checkpoint predates target selection,\nrun the six steps above. Changing an upstream selection resets all downstream\ntracker values.\n\nTreat unresolved target configuration as a blocking interactive state, not a\nsuccessful stopping point. After the export summary, immediately ask for the\nservice type; after each answer, ask for the dependent objective and vCore value\nuntil `Complete-TargetConfiguration` succeeds. Do not end the run, report the\nmigration as complete, or leave imports merely `Pending` while\n`TargetConfiguration` is null unless the user explicitly pauses or cancels.\n\nAfter the modern `sqlcmd` gate passes, default target import authentication to\ninteractive Microsoft Entra ID. Do not ask the user to choose an authentication\nmethod unless they explicitly request the default credential chain. When choices\nmust be shown, present them in this order:\n\n| Target authentication | Command | Recommendation |\n| --- | --- | --- |\n| Interactive Microsoft Entra ID | `sqlcmd -S \u003Cserver> -G -U \u003Cuser>` | **Default**; attended and supports MFA |\n| Microsoft Entra default credential chain (`ActiveDirectoryDefault`) | `sqlcmd -S \u003Cserver> --authentication-method ActiveDirectoryDefault` | Optional explicit override when an Azure CLI, developer credential, or managed identity should be reused |\n\nFor interactive Entra ID, require the user's UPN and never place a password or\ntoken on the command line. Reject SQL login authentication; never invoke a password\nprompt or read credentials from Key Vault, Credential Manager, or a masked control.\nUse the approval-time target preflight as the authentication\u002Fconnectivity check;\ndo not launch a separate preliminary `sqlcmd` probe. After that preflight succeeds,\ncall `New-TargetConnectionStringForApprovedAuthentication` to build the equivalent\nsecretless, `SqlPackage`-supported target connection in memory without another\nbrowser prompt. Do not read an ambient `$targetConnectionString` variable, and do\nnot display or persist the generated value. `SqlPackage` authenticates independently\nwith the selected method; no `sqlcmd` token is handed off. Use\n`InteractiveEntra` unless the user explicitly selects `ActiveDirectoryDefault`.\nSeparate `sqlcmd` and `SqlPackage` processes can each display an interactive\nsign-in prompt.\nThen show the exact Phase 4 approved Azure SQL Database settings for the batch:\n\n- service type (`GeneralPurpose`, `BusinessCritical`, or `Hyperscale`);\n- approved catalog service objective;\n- explicitly selected vCore count;\n- derived SqlPackage database service objective;\n- automatically applied catalog maximum database size in GB.\n\nDo not change SKU settings after the Phase 4 selection is checkpointed. Revalidate\nthe immutable selection against the same approved catalog before displaying it.\nThe single\ngrouped approval must be tabular;\ndo not describe the eligible databases, BACPAC paths, or target settings in a prose\nparagraph. Render these two tables before requesting approval:\n\n| DatabaseName | ExportStatus | BacpacPath |\n| --- | --- | --- |\n| `\u003Ctarget database>` | `Success` | `\u003Cfull local BACPAC path>` |\n\n| TargetServer | Authentication | ServiceType | ServiceObjective | VCore | DatabaseServiceObjective | MaximumSizeGB | Overwrite |\n| --- | --- | --- | --- | ---: | --- | ---: | --- |\n| `\u003Clogical server>` | `\u003Cmethod>` | `\u003CGeneralPurpose \u002F BusinessCritical \u002F Hyperscale>` | `\u003Ccatalog objective>` | `\u003CvCore>` | `\u003Cderived objective>` | `\u003Csize>` | `No` |\n\nAfter authentication, use that same approved identity and validated Go `sqlcmd`\nroute for one injection-safe, read-only `sys.databases` query covering every\nsuccessfully exported target name. Run this preflight before displaying the\napproval tables.\nMark every name observed by this query as `Skipped - database existed before\nimport`, exclude it from the approval table, and retain the observation as\nevidence. If the preflight cannot complete, stop before approval and do not import;\ndo not fall back to diagnostic-text matching.\n\nAfter approval, run exactly one final pre-import preflight and consume its\nscoped `FinalImport` receipt directly in the import loop. There are exactly two\ntarget preflight calls in a successful run: the approval-time call and the final\npre-import call. Never add a third preflight inside a generated import script.\nEach preflight must have a bounded timeout; stop the exact `sqlcmd` process and\npause the migration if it does not return within the configured limit.\nUse a 600-second default for interactive Entra browser\u002FMFA authentication and\n180 seconds for `ActiveDirectoryDefault`. Keep interactive `sqlcmd` attached to\nthe current console and use its `-o` option for structured result capture; do\nnot redirect standard input, output, or error while the attended login runs.\n\nThe final preflight is a point-in-time collision check, not a reservation or\nownership guarantee. With SqlPackage-only target mutation, another actor can\ncreate the name after preflight and before or during import. On success, report\n`TargetStateAfterImport = ImportedSuccessfully`; do not claim that the workflow\ncreated or owns the Azure resource.\n\nAfter the tables, ask one short approval question that states the import is offline\nand authorizes `SqlPackage.exe \u002FAction:Import` against the listed target database\nnames with this residual race disclosed. Do not repeat the table contents in\nprose. After approval, use only\n`SqlPackage.exe \u002FAction:Import` for target mutation and do not call an Azure API.\n\nImport only entries whose `ExportStatus` is `Succeeded`; for each, require its\nselected folder and a file named `\u003Csubfolder-name>.bacpac`. Do not treat a failed\nexport's folder as an import error.\n\nImmediately before each import, validate the BACPAC at the exact manifest path:\n\n- Confirm the expected BACPAC exists and has a plausible nonzero size.\n- Run the BACPAC verification\u002Fchecksum process used by the operator.\n- Do not import when either check fails. Mark that import `Failed` and\n  `AwaitingManualRemediation`, preserve the artifact and diagnostics, and continue\n  only with unaffected databases when `SelectionMode` is `All`.\n\nUse the local import loop in section 4 of\n[references\u002Fcommand-execution-export-import.md](references\u002Fcommand-execution-export-import.md).\n\nUse one `\u002FTargetConnectionString` containing the selected supported authentication\nmethod, server, and database name. Never print that argument. Pass the SKU with\n`\u002Fp:DatabaseEdition`, `\u002Fp:DatabaseServiceObjective`, and\n`\u002Fp:DatabaseMaximumSize`. Do not mix `\u002FTargetConnectionString` with other target\nconnection parameters, including `\u002FTargetDatabaseName`. When constructing the\nper-database connection string with `DbConnectionStringBuilder`, invoke its\n`set_ConnectionString()` and `get_ConnectionString()` methods explicitly;\nPowerShell property assignment can be routed through the builder's dictionary\ninterface and create an invalid nested `ConnectionString` key.\n\nBefore launching the first import, state that live progress will appear in the\nterminal every 30 seconds. Print `Started` immediately for every import, including\nits position in the batch, then poll every 30 seconds and report the database,\nelapsed time, process state, diagnostics-file size and growth, and latest sanitized\ndiagnostic status. Use `Running` when diagnostics changed, `Running - no new\ndiagnostics yet` while the process remains active, and `Possibly stalled` after\nfive minutes without diagnostic activity. The latter is informational and must not\nstop the process before the configured timeout. Import has no growing local output\nfile equivalent to the export BACPAC, so use process state and diagnostics growth\nas movement indicators without claiming a percentage. If an active import reports\n`Incorrect syntax near 'EDITION'` while the target database is already visible as\n`ONLINE`, treat it as a nonterminal SqlPackage diagnostic-masking\u002Fparser message:\nreport it as still running and continue polling until the SqlPackage process exits.\nDo not classify that text alone as a failure, do not restart the import, and do not\nswitch authentication or SKU settings because of it. Do not leave a failed\noperation in `InProgress` or continue polling an exited process. Persist the same\nsanitized lines to the current attempt's\n`\u003Cdatabase-folder>\\attempts\\import\\\u003Cattempt-id>\\status.log`; when an agent or\ncommand runner cannot stream a blocking command's output, it must tail this file\nand relay the latest heartbeat rather than leaving the user without an update.\n\nImport gates:\n\n- Record the `SqlPackage` exit result for every attempted database; only mark an\n  individual import `Succeeded` when its exit code confirms success.\n- Record import duration and any warnings.\n- Stop if the import targets an unexpected server or database.\n\nFor an `All` selection, process only the folders recorded in the manifest. A\ndatabase may be marked `Skipped` only by the successful pre-import\n`sys.databases` observation. For every nonzero SqlPackage exit, record its\nsanitized reason and diagnostics path, mark it `Failed`, and state that the target\nmay have been partially created and requires explicit inspection before retry.\nNever convert an attempted import failure to `Skipped` by matching diagnostic\ntext, regardless of wording or localization. Re-run the same authoritative\ndatabase-existence query after a failed import and record whether the target is\npresent or absent; if that inspection fails, stop the batch rather than continuing\nwith unknown target state. Continue importing the remaining BACPACs when `All` is\nselected only after that inspection succeeds. After all import attempts, display one tabular row per\nattempted database with `DatabaseName`, `Status` (`Success`, `Skipped`, `Fail`, or\n`InProgress`), `ServiceType`, `ServiceObjective`, `VCore`,\n`DatabaseServiceObjective`, and `MaximumSizeGB`. Keep failure\nreasons and diagnostics paths below the table. Invoke validation only for entries\nwhose `ImportStatus` is `Succeeded`.\n\nAfter each database reaches a terminal state, also display the current batch as a\ntable with `DatabaseName`, `Status`, `Elapsed`, and `Detail`. Rows not yet attempted\nmust remain `Pending`; the active row is `InProgress`; completed rows are `Success`,\n`Skipped`, or `Fail`. This intermediate table supplements rather than replaces the\n30-second heartbeat and final import summary.\n\n## Phase 5 — Hand off post-migration validation\n\nFor each manifest entry whose `ImportStatus` is `Succeeded`, invoke\n`validate-post-migration-data` with the existing source connection, target\nconnection details and authentication route, database mapping, source write-freeze\nevidence, approved Azure SQL Database adaptations, and evidence location. Do not\ninvoke validation for skipped exports or imports. Do not request credentials again\nor run data\u002Fschema reconciliation in this migration skill.\n\nStore only the returned validation report status and location in the migration\nmanifest. Do not mark the overall migration complete when any selected database\nreport is `Fail` or `In progress`. This workflow does not delete the source\ndatabase, BACPAC files, diagnostics, manifest, or validation evidence. Cleanup is\na separate, explicitly requested process.\n\n\n\n\n## Operational failures and routing\n\nDo not proactively scan the source for these conditions. Use this table only when\nan export or import command reports the corresponding failure.\n\n| Failure | Action |\n| --- | --- |\n| BACPAC destination file already exists | Do not start or continue that export and do not overwrite, rename, delete, or reuse the file. Mark the item `Failed` and `AwaitingManualRemediation`, save the checkpoint, report the exact path and name collision, and direct the user to preserve\u002Fmove the existing evidence or choose a new export root externally. Resume after `source remediation complete` only when the intended destination no longer exists. |\n| Unsupported `SqlPackage` argument | Stop the batch, mark affected items `AwaitingManualRemediation`, save the checkpoint, and report the argument, installed SqlPackage version, diagnostics, and manual template\u002Fconfiguration correction required. Do not edit or rerun the command inline. Resume failed\u002Fpending databases only after user confirmation. |\n| Unsupported source object or feature | Mark the database `Failed` and `AwaitingManualRemediation`, save the sanitized checkpoint, and provide diagnostics plus external assessment\u002Fchange\u002Ftest\u002Fresume steps. Do not change source code\u002Fschema, restore a substitute database, or invoke remediation inline. Stop for a single database; for `All`, continue only unaffected databases. |\n| Authentication failure during export or import | Mark the database `Failed` and `AwaitingManualRemediation`, save the sanitized checkpoint, and provide identity, permission, network\u002Ffirewall, secure-store or MFA verification steps appropriate to the selected method. Do not retry, refresh, or switch authentication inline. |\n| Target database exists before import | Observe it with the injection-safe, read-only `sys.databases` preflight through validated Go `sqlcmd`, mark it `Skipped - database existed before import`, and exclude it from import approval\u002Fexecution. A SqlPackage failure never proves this condition. |\n| Target may have been created by a failed import | Keep the attempted import `Failed`, preserve diagnostics, and require explicit target inspection before retry. Do not relabel it `Skipped` merely because a database now exists. |\n| `Incorrect syntax near 'EDITION'` appears while import is still running and target database is `ONLINE` | Treat as nonterminal SqlPackage diagnostic-masking\u002Fparser output. Continue polling the active process and use only process exit code plus authoritative target inspection after exit for terminal classification. |\n| Import timeout or throttling | Mark the import `Failed` and `AwaitingManualRemediation`, save the checkpoint, and instruct the user to address target sizing, throttling, or connectivity externally. Retry the same verified BACPAC only after `target remediation complete`. |\n| Missing target user or permission | Route identity remediation to the target security workflow. |\n| Target capacity is insufficient | Stop and route to the sizing\u002Fplanning skill before retrying import. |\n\n## Manual-remediation resume gate\n\nPersist a sanitized checkpoint after every terminal export\u002Fimport result. It may\ncontain nonsecret mappings, statuses, timestamps, diagnostics paths, failure\ncategories, and manual next actions, but never connection strings, passwords,\ntokens, or command lines containing them. On resume, treat the checkpoint only as\na cursor: revalidate the source\u002Ftarget identity, selected database mapping, local\nBACPAC state, and target preflight from authoritative sources.\n\nRetry an `AwaitingManualRemediation` item only after the user explicitly states\n`authentication remediation complete`, `source remediation complete`, or `target\nremediation complete` as applicable. Display the failed database, prior sanitized\nreason, and intended retry before proceeding. Preserve successful exports\u002Fimports\nand do not repeat them. If the user instead requests a fresh execution, rebuild\nthe manifest after discovery in a new export root. Never overwrite an existing\ncheckpoint, BACPAC, diagnostics file, or status file during a fresh execution;\nretain the old export root as immutable evidence.\n\nEvery export or import invocation, including an approved retry, must create a new\nUTC-and-GUID attempt directory. Record each attempt ID, timestamps, result, and\nits diagnostics, console, and status paths in the sanitized checkpoint. Never\nreuse or truncate an earlier attempt's evidence files.\n\n## Result format\n\nAfter planning, report:\n\n> **Migration plan — `\u003Csource database or all selectable user databases>`**\n> **Azure SQL Database** via **BACPAC** · downtime **offline**\n\n| | Decision |\n| --- | --- |\n| Target | Azure SQL Database — `\u003Clogical server name>` |\n| Method | BACPAC export\u002Fimport with `SqlPackage` |\n| Import execution | `SqlPackage.exe \u002FAction:Import` only |\n| Completion | `\u003CComplete \u002F Failed \u002F In progress>` |\n\nInclude:\n\n- Selected database scope and resolved export root.\n- One local BACPAC path per database.\n- Per-database export\u002Fimport operation status.\n- Last status-update time, sanitized failure reason, and diagnostics location for\n  every failed or long-running operation.\n- Per-database validation handoff\u002Freport status and report location.\n\nFor a failed operation, state the failed phase, error category, evidence\nlocation, and the remediation required. Never present a partial migration as\ncomplete.\n\n## References\n\n- [Local BACPAC command execution — manifest setup](references\u002Fcommand-execution-setup.md)\n- [Local BACPAC command execution — export and import](references\u002Fcommand-execution-export-import.md)\n- [SqlPackage export action](https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Fsql\u002Ftools\u002Fsqlpackage\u002Fsqlpackage-export)\n- [SqlPackage import action](https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Fsql\u002Ftools\u002Fsqlpackage\u002Fsqlpackage-import)\n- [Azure SQL Database migration overview](https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Fdata-migration\u002Fsql-server\u002Fdatabase\u002Foverview)\n- [Azure SQL Database elastic jobs](https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Fazure\u002Fazure-sql\u002Fdatabase\u002Felastic-jobs-overview)\n","# SQL Server to Azure SQL Database using BACPAC\n\n## When to Use\n\n- Migrate one or more SQL Server user databases to Azure SQL Database using BACPAC export and import.\n- Perform an offline Azure SQL Database migration with `SqlPackage`.\n\nHelp a user migrate one or all user databases from **SQL Server** to **Azure SQL\nDatabase** using an offline **BACPAC** workflow. A BACPAC contains the database schema and data. It\ndoes not preserve instance-level objects, server configuration, SQL Agent jobs,\nor a transactionally consistent online migration.\n\nUse this skill when the user asks to migrate a SQL Server database to Azure SQL\nDatabase with BACPAC, `SqlPackage`, import\u002Fexport, or an offline database move.\n\nRead [references\u002Fcommand-execution-setup.md](references\u002Fcommand-execution-setup.md) and\n[references\u002Fcommand-execution-export-import.md](references\u002Fcommand-execution-export-import.md)\nbefore generating or running commands.\nResolve the absolute installed directory containing those reference files and\npass it to the command blocks as `$skillReferenceRoot`. Do not derive that path\nfrom the current directory or `$PSScriptRoot`; the orchestration runs inline,\nnot from a script file.\n\nThe `Invoke-SqlPackageWithProgress` implementation in the checked-in setup\nreference is the only permitted process boundary for `SqlPackage` export and\nimport. Do not replace it with `Start-Process`, a shell command string, an\nalternate script, or an executor-built wrapper. It uses\n`ProcessStartInfo.ArgumentList` so every complete `\u002FName:\u003Cvalue>` element remains\none operating-system process argument and rejects empty or detached\nconnection-string and BACPAC-path fragments before launch.\nUse only the manifest, checkpoint, and per-database folder layout defined by the\nreference blocks; do not introduce aggregate `exports`, `manifest`, `evidence`,\nor `checkpoint` folders.\n\nValidate modern Go `sqlcmd` capabilities from the combined output of every\nsuccessful supported help mode (`--help` and `-?`). A required flag may appear\nin either output. Report the successful help modes and exact missing flags before\noffering installation; do not treat a reduced `--help` view as proof that an\ninstalled client lacks flags exposed by `-?`. Quote the compatibility argument as\n`'-?'`, and name wrapper parameters `$ArgumentList`, never `$Args`, so PowerShell's\nautomatic `$args` variable cannot consume the intended arguments. After an\napproved `winget install sqlcmd`, rerun capability validation regardless of the\nwinget exit code; an already-installed\u002Fno-upgrade result is successful when the\nrefreshed client passes validation.\n\n\n## Operating contract\n\n- Collect all missing nonsecret migration inputs in one grouped request whenever\n  possible, except database selection must follow authoritative source discovery\n  when the user has not already supplied an exact database name or `All`. Reuse\n  values already supplied by the user or discovered safely.\n- Treat this as an **offline migration**. State that downtime will occur, but do\n  not request a separate downtime acknowledgment. Include that acknowledgment in\n  the single target-database creation approval. Do not ask when downtime should\n  start, how long it may last, or what duration is acceptable.\n- Require the user to select one online user database or all selectable online\n  user databases. Never interpret `all` as including system databases.\n- Do not run source migration readiness, compatibility, feature, dependency,\n  performance, or sizing assessments. Even if the database is online and in read-write mode, allow the export to continue. However, inform the user that they must ensure there is no DML  and DDL activity on the database during the export.Consume prior skill outputs when supplied;\n  otherwise use the source and target decisions provided by the user.\n- For target details, require the existing Azure SQL logical server name and an\n  explicit target configuration after export: service type (`GeneralPurpose`,\n  `BusinessCritical`, or `Hyperscale`), service objective, and vCore count. Apply\n  the catalog's documented maximum size automatically. Before export, collect only the existing logical server\n  name. After every export is attempted and at least one succeeds, track each\n  user input separately. Never infer a representative objective or vCore count\n  from the service type. Validate the resolved configuration with\n  `Resolve-ApprovedTargetSkuSelection`; do not accept an arbitrary objective or\n  size outside the catalog. Do not ask for subscription, resource group, region, or target\n  authentication before export.\n- Never connect to either the source or target with SQL login\u002FSQL authentication.\n  Do not request, accept, retrieve, or use SQL passwords or SQL credentials. Use\n  only Windows Integrated authentication or Microsoft Entra authentication\n  supported by the workflow. Never read SQL credentials from Key Vault or\n  Credential Manager, or expose, print, commit, or persist them.\n- Run each `SqlPackage` operation with bounded status polling. Emit a progress\n  update every 30 seconds with the database, batch position, operation, elapsed time, process\n  state, diagnostics\u002FBACPAC file growth, and latest sanitized diagnostic status\n  so the user can distinguish movement from a potentially stalled operation.\n  Write each static heartbeat to both the live terminal and a sanitized\n  operation-status file beside the diagnostics, and flush host output after each\n  line. If the command runner buffers terminal output until completion, tail the\n  status file and relay its latest line at least every 30 seconds; do not rely on\n  the transient `Write-Progress` display as the only visible progress channel.\n  Print an immediate `Started` message before the first poll and tell the user\n  that live status appears in the terminal. After five minutes without diagnostic\n  or BACPAC growth, report `Possibly stalled` while making clear that the process\n  is still active. Do not invent a percentage when `SqlPackage` does not provide one.\n- Pass complete `\u002FSourceConnectionString:\u003Cvalue>`,\n  `\u002FTargetConnectionString:\u003Cvalue>`, `\u002FSourceFile:\u003Cvalue>`, and\n  `\u002FTargetFile:\u003Cvalue>` forms as individual string-array elements to the\n  checked-in progress wrapper. Never split a switch name from its value or\n  flatten the argument array into a command string.\n- Do not execute destructive operations against the source database.\n- After all exports and target authentication, run one injection-safe, read-only\n  target `sys.databases` preflight through validated Go `sqlcmd` for the selected\n  names. Mark databases observed\n  before import as skipped. Never infer this condition from localized SqlPackage\n  diagnostics.\n  - Do not perform any write or edit or perform any DDL opertion in the source databases\n- Surface errors and record the failed phase. For an `All` export or import, skip\n  the failed database and continue the remaining databases. Do not report success\n  based on a command that was not confirmed successful.\n- Use Windows PowerShell syntax when running local commands on Windows.\n\n## Assessment boundary\n\nThis skill assumes a separate assessment or user decision has already selected\nAzure SQL Database, BACPAC, and any required source remediation.\nDo not inspect the source for migration blockers or reevaluate the selected target\nand method. If a prerequisite is missing, route to the appropriate assessment or\nplanning skill rather than adding assessment prompts to this workflow.\n\nAn export error may identify an unsupported object or feature. Record the error\nand skip that database when processing an `All` selection; do not broaden the\nmigration run into a source assessment.\n\n\n## Workflow\n\n```text\nPhase 0: Validate tooling, then collect source connection and safety inputs\nPhase 1: Discover databases, resolve source scope, and prepare the complete execution manifest\nPhase 2: Prepare local folders without an approval prompt\nPhase 3: Export each selected database to its BACPAC subfolder\nPhase 4: Select target SKU and size, authenticate, approve, then import with SqlPackage\nPhase 5: Hand every successfully imported database to post-migration validation\n```\n\n## Phase 0 — Collect inputs and safety gates\n\nTreat modern Go-based `sqlcmd` as a mandatory prerequisite for the BACPAC\nworkflow. Before discovery, folder creation, or export, inspect every `sqlcmd`\nexecutable on `PATH`; do not assume the first result is modern because ODBC and\nGo-based versions can coexist. Select only a candidate whose `--version` command\nsucceeds and whose help exposes `-G`, `-U`, and `--authentication-method`. If no\ncandidate passes, ask once for approval to install modern `sqlcmd` with the official\n`winget install sqlcmd` command. Do not install silently. If approved, install it,\nrefresh the current process `PATH`, and repeat the checks. If installation is\ndeclined or validation still fails, stop before starting the migration and provide\nthe official manual-download link. Cache the resolved full path for Phase 4.\n\nDo not ask for target service type, objective, vCore, or maximum size in Phase 0.\nAfter Phase 3 has attempted every selected export and at least one succeeds, load\nthe checked-in catalog\n`references\u002Fazure-sql-db-vcore-sku-limits.json` with\n`Get-ApprovedTargetSkuCatalog` from `references\u002Ftarget-sku-validation.ps1`.\nThe catalog is derived from Microsoft Learn's *Single database vCore resource\nlimits* page, source-dated March 9, 2026 and updated April 10, 2026. Use the\nstaged input tracker and make collection interactive:\n\n1. Present exactly `General Purpose`, `Business Critical`, and `Hyperscale` as\n  service-type selections.\n2. After that selection, filter the approved catalog and present only its exact\n  provisioned\u002Fserverless hardware objective labels for the chosen type. Do not\n  present a DTU `Standard`\u002F`Premium` tier as a vCore objective.\n3. After the objective selection, load its approved vCore values and require an\n  explicit integer selection\u002Finput.\n4. Then display the documented maximum data size for that exact vCore and apply\n  it automatically without a separate confirmation prompt. Catalog entries must provide a\n  `DocumentedMaxDataSizeGBByVCore` map. It records the resource-limit ceiling,\n  not every configurable storage increment.\n\nKeep separate status\u002Fvalue\u002Foptions entries for `ServiceType`,\n`ServiceObjective`, `VCore`, and `MaximumSizeGB`. Changing service type clears\nthe objective, vCore, and size; changing objective clears vCore and size. Never\nselect the first option or infer `1024 GB` for Hyperscale. The approved catalog\nmust map the friendly objective and vCore to an exact\n`DatabaseServiceObjectiveTemplate` used to derive the SqlPackage objective\n(for example `HS_Gen5_{0}` plus vCore `4` becomes `HS_Gen5_4`). Validate the\ncomplete selection through `Resolve-ApprovedTargetSkuSelection`. Do not use\nSqlPackage import to discover that a SKU is unavailable for the target server,\nregion, subscription, quota, or capacity.\n\nThe checked-in stable SKU list is:\n\n| Service type | Compute\u002Fhardware objective | Published vCore choices | Published max data size |\n| --- | --- | --- | --- |\n| General Purpose | Provisioned standard-series (Gen5) | 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 24, 32, 40, 80, 128 | 1–4 TB, depending on vCore |\n| General Purpose | Serverless standard-series (Gen5) | 1, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 24, 32, 40, 80 max vCores | 512 GB–4 TB, depending on max vCore |\n| General Purpose | Provisioned DC-series | 2, 4, 6, 8 | 1–3 TB, depending on vCore |\n| Business Critical | Provisioned standard-series (Gen5) | 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 24, 32, 40, 80, 128 | 1–4 TB, depending on vCore |\n| Business Critical | Provisioned DC-series | 2, 4, 6, 8 | 768 GB |\n| Hyperscale | Provisioned standard-series (Gen5) | 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 24, 32, 40, 80 | 128 TB |\n| Hyperscale | Serverless standard-series (Gen5) | 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 24, 32, 40, 80 max vCores | 128 TB |\n| Hyperscale | Provisioned DC-series | 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 32, 40 | 128 TB |\n| Hyperscale | Provisioned premium-series | 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 24, 32, 40, 64, 80, 128 | 128 TB |\n| Hyperscale | Provisioned premium-series memory optimized | 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 24, 32, 40, 64, 80 | 128 TB |\n\nExclude Fsv2-series because it is unavailable for new creation and scheduled for\nretirement on October 1, 2026. Exclude preview choices, including General Purpose\nand Business Critical DC-series 10–40 vCores and Hyperscale premium-series 160\nand 192 vCores. Treat every availability note as a guardrail: the catalog does\nnot prove regional availability, quota, subscription eligibility, or capacity.\n\nProduce a concise migration plan before execution:\n\n| Item | Value |\n| --- | --- |\n| Source Server | `\u003Cexisting SQL Server instance name>` |\n| Target server | `\u003Cexisting Azure SQL logical server name>` |\n| Target configuration | `Selected after export` |\n| Method | Local BACPAC export\u002Fimport with `SqlPackage` |\n| Downtime | Offline; acknowledged in the target-database creation approval |\n| Overwrite behavior | Create new database; never overwrite by default |\n| Local root | `\u003Cuser-provided path or default path>` |\n\nCollect the missing source authentication, existing target logical server, and\nexport root in one grouped request. Source authentication is never inferred from\nthe host or silently defaulted: require an explicit selection of `Windows\nIntegrated` or `Microsoft Entra Interactive MFA`. For the export root, show\n`$env:USERPROFILE\\SqlMigration\\Bacpac` as the recommended default and require the\nuser either to accept it or provide another absolute path; never append a\ntimestamp or create a run directory before this choice is resolved. Reuse a value\nonly when the user supplied it explicitly in the current request or earlier in\nthe conversation.\n\nBefore discovery or any filesystem write, create an in-memory input-resolution\nrecord containing `SourceServer`, `SourceAuthentication`, `TargetServer`, and\n`ExportRoot`. If any field is unresolved, stop and ask the grouped question. Do\nnot run discovery, create directories or checkpoints, or invoke `SqlPackage`\nuntil all four fields are resolved. Echo the resolved nonsecret values to the\nuser before proceeding.\n\nIf the user's original request already names one exact source database or says\n`All`, retain that value as the requested selection and do not ask for it again\nbefore discovery. Otherwise, defer the database-selection field until Phase 1 so\nit can contain the discovered database names. Input collection is not execution\napproval. Do not request approval in Phases 0-3, and do not collect a downtime\nstart time, window, maximum duration, RPO, or RTO.\n\n## Phase 1 — Discover databases and build the manifest\n\nConnect to the source database using the following authentication methods:\na. Windows Authentication\nb. Entra Authentication using Interactive MFA sign-in\n\nAfter source authentication is established, always discover all online user\ndatabases from the source instance before presenting a database-selection form.\nExclude `master`, `model`, `msdb`, `tempdb`, snapshots, and offline databases.\nThis is scope discovery, not migration assessment.\n\nResolve the database scope as follows:\n\n1. If the user's original request or prior answer already supplied one exact\n   database name, match it against the discovered selectable list. When it\n   matches, reuse it without asking again. When it does not match, show the\n   discovered list and ask for a corrected selection.\n2. If the user already supplied `All`, reuse it without asking again; `All`\n   means every database in the discovered selectable list.\n3. Otherwise, display a single-choice form whose options are `All` followed by\n   every exact discovered database name. Never display `All` as the only preset\n   when one or more selectable databases were discovered, and do not require the\n   user to type a discovered database name as free text.\n\nIf no selectable online user databases are discovered, stop with a clear error\ninstead of showing an empty or `All`-only selection form.\n\n\nRequire the source authentication selection made in the Phase 0 grouped request.\nIf Microsoft Entra authentication is selected, use the supported `SqlPackage`\nauthentication options for the installed version.\nReject SQL login authentication and do not request or use a password from any\nsource, including Key Vault, Credential Manager, masked controls, or `sqlcmd`.\n\nResolve the export root from the grouped input request. The recommended default is\n`$env:USERPROFILE\\SqlMigration\\Bacpac`, but it is not selected until the user\naccepts it. Use the accepted path exactly as the export root; do not silently add\na timestamped child folder. Build the complete in-memory manifest and database\nmappings without creating folders, databases, or BACPAC files.\n\n\n## Phase 2 — Prepare without prompting\n\nDisplay the final source scope, target server name, database mappings, local root,\nand overwrite behavior. State clearly that the workflow is offline and will cause\ndowntime.\nDo not request execution confirmation here. Prepare local folders and run every\nselected export from the displayed manifest. Do not prompt between databases,\nfor authentication retry routing, after failures, or between export and import.\nAn authentication or unsupported-object failure is not remediated inline: save\nthe database as awaiting manual remediation, give concrete external next steps,\nand continue only with unaffected databases when `All` was selected.\n\nStop with a clear error when required nonsecret input is missing or inconsistent.\nChanging application connection strings remains a separate cutover action.\n\nThis skill does not discover, create, or configure Azure resources. The target\nlogical server must already exist. Do not open a target SQL data-plane connection,\nrequest target credentials, or test target database authentication before every\nselected export has been attempted. The import phase creates one database for each\nsuccessfully exported source database after confirming that the name is not in use.\n\n### Prepare local folders\n\nFor every selected database, use this exact layout:\n\n```text\n\u003Cexport-root>\\\n  \u003Cdatabase-name>\\\n  \u003Cdatabase-name>.bacpac\n  attempts\\\n    export\\\u003CUTC-GUID-attempt-id>\\\n      diagnostics.log\n      console.log\n      status.log\n    import\\\u003CUTC-GUID-attempt-id>\\\n      diagnostics.log\n      console.log\n      status.log\n```\n\nThe database name must be valid as a Windows folder name. If it contains an\ninvalid path character, is a reserved Windows name, would resolve outside the\nexport root, or maps to an existing symbolic link or junction, exclude that\ndatabase and display a warning that names the database and the reason it was\nskipped. Do not silently sanitize it because the subfolder name controls the\ntarget database name. If no eligible databases remain, stop before creating\nfolders.\n\nPersist the user's original database selection as `SelectionMode`, with the\nvalue `All` or `Explicit`; never infer this mode from the number of manifest\nentries after discovery or path filtering. The migration manifest contains\n`SourceDatabase`, `FolderPath`, `BacpacPath`,\n`TargetDatabase`, `ExportStatus`, `ExportFailureReason`, `ExportLastUpdatedUtc`,\n`ImportStatus`, `ImportFailureReason`, `ImportLastUpdatedUtc`,\n`ImportStartedUtc`, `ImportCompletedUtc`, `ImportDuration`,\n`ImportLastActivityUtc`, `TargetStateAfterImport`, `ResumeState`,\n`FailureCategory`, `ManualNextAction`, and\n`ValidationReportStatus`. Set `TargetDatabase` to the subfolder name. Use this\nsame manifest through export, import, and the validation handoff; do not rediscover\nor broaden the selection later.\n\nUse section 2 of [references\u002Fcommand-execution-setup.md](references\u002Fcommand-execution-setup.md) after the\nmanifest is displayed; no approval is required to create local folders.\n\n## Phase 3 — Export with SqlPackage\n\nUse the latest supported `SqlPackage` installed on the operator machine.\nValidate the executable and version before running the export.\n\nUse the export loop in section 3 of\n[references\u002Fcommand-execution-export-import.md](references\u002Fcommand-execution-export-import.md) and obtain connection\ndetails from an approved secure provider.\n\nTreat the authenticated source connection as an immutable batch-level base. For\nevery manifest row, create a fresh connection-string builder, remove both\n`Database` and `Initial Catalog` aliases, and set `Initial Catalog` to that row's\n`SourceDatabase` unconditionally. A catalog supplied by the base connection,\nincluding `master`, must never determine which database is exported.\n\nSet the manifest status to `InProgress` before starting `SqlPackage`. Poll the\nprocess every 30 seconds, update `ExportLastUpdatedUtc`, and display a console\nheartbeat without exposing arguments or credentials. Include whether diagnostic\nor BACPAC bytes increased since the previous poll and the latest sanitized\ndiagnostic status line. File growth is evidence of movement, not a completion\npercentage; if neither file grows, state that the process is still running rather\nthan claiming it is stuck.\n\nWhen the process exits, display `Succeeded` or `Failed`\nimmediately. On failure, include the database, exit code, sanitized reason, and\ndiagnostics path in the user-facing update.\n\nFor Microsoft Entra authentication, use the supported `SqlPackage`\nauthentication options for the installed version. Do not put an access token in\nthe command line.\n\nFor Windows integrated authentication, place `Integrated Security=True` (or\n`Trusted_Connection=True`) inside `\u002FSourceConnectionString`. Never generate\n`\u002FSourceIntegratedSecurity`; it is not a supported `SqlPackage` 170.4 argument.\nValidate the complete argument template before the first database export so an\nunsupported switch cannot fail multiple databases. If an already-running batch\nreports an unrecognized argument, stop that batch, report the configuration\nfailure immediately, save the checkpoint, and provide the installed SqlPackage\nversion plus the unsupported argument as manual correction guidance. Do not edit\nor rerun the command inline. Resume only failed\u002Fpending databases after the user\nconfirms the external configuration correction; do not re-export successful ones.\nTreat configuration correction as manual remediation; the confirmation phrase is\n`source remediation complete` because it unblocks the export stage, even though\nno source database change is authorized.\n\nExport gates:\n\n- Immediately before each export, check the exact manifest `BacpacPath` with a\n  literal-path file test. If a file already exists at that path, do not start\n  SqlPackage and never overwrite, rename, delete, or reuse the file. Mark that\n  export `Failed` and `AwaitingManualRemediation`, report the full path and\n  `A BACPAC file with the same name already exists`, save the checkpoint, and\n  instruct the user to preserve\u002Fmove the existing evidence or choose a new export\n  root outside this workflow. Resume only after `source remediation complete`\n  and after confirming that the intended path no longer exists. For `All`,\n  continue only with databases whose BACPAC destination does not already exist.\n- Confirm the export exit code is successful.\n- Confirm the export produced the expected BACPAC file. Artifact validation is\n  deferred until immediately before import.\n- Confirm the source was not modified by the export workflow.\n- Record the export timestamp and source database state.\n\nIf export fails, preserve the diagnostics and record the database and failure\nreason. Never silently omit a failed database or failed object from the report.\n\nIf diagnostics identify an authentication failure, do not refresh credentials,\nretry, prompt for another authentication method, or switch to an Entra\u002FSQL\u002Fmanaged\nidentity fallback in the current execution. Mark the export `Failed`, set\n`ResumeState = AwaitingManualRemediation`, save a sanitized checkpoint, and show:\n\n1. the selected authentication method and sanitized SqlPackage error;\n2. the diagnostics and checkpoint locations;\n3. instructions to verify the same identity can connect to and read the source\n  database, correct its login\u002Fuser permissions, firewall\u002Fnetwork route, MFA or\n  approved secure-store entry outside this workflow as applicable; and\n4. instructions to return with `authentication remediation complete` to resume\n  the failed database, or restart with the same export root.\n\nNever request, print, or persist a password or token while reporting the failure.\nDo not retry until the user confirms that external remediation is complete; then\nrevalidate the connection and retry only the failed\u002Fpending database. A different\nauthentication method is a restarted\u002Frevised run, not inline remediation.\n\nIf diagnostics identify unsupported source schema, code, or objects, do not alter\nDDL, drop\u002Fdisable an object, restore or clone another database, export from a\nsubstitute database, or invoke an automated remediation. Record the exact\nunsupported item when diagnostics provide it, set the same awaiting-remediation\nstate, and direct the user to review the diagnostics, assess and approve the\nrequired source change outside this skill, back up the source, apply and test that\nchange manually, and return with `source remediation complete`. Do not invent a\ngeneric schema change when SqlPackage does not identify one.\n\nFor an `All` selection, mark that database `Failed`, skip it, and continue exporting\nthe remaining databases. After every selected export has been attempted, proceed\nwith only manifest items whose `ExportStatus` is `Succeeded`. If no exports\nsucceeded, stop before target connectivity. For a single-database selection, stop\nwhen its export fails. After the export loop, display a summary of succeeded and\nfailed databases as a table with `DatabaseName`, `Status` (`Success`, `Fail`, or\n`InProgress`), and human-readable `BacpacSize`. Show `Not created` when no BACPAC\nfile exists. Include each failed database's sanitized reason and diagnostics path\nbefore starting target connectivity.\n\n## Phase 4 — Initiate target connectivity and import\n\nAfter every selected export has reached `Succeeded` or `Failed` and at least one\nBACPAC succeeded, reuse the modern Go-based `sqlcmd` executable validated in\nPhase 0 and invoke it by its resolved full path. If that exact executable is no\nlonger available, stop before target connectivity and report that the prerequisite\nchanged during the run; do not substitute an unvalidated executable.\n\nBefore target authentication or connectivity, select and freeze the target SKU\nand size using these steps:\n\n1. Load the checked-in catalog with `Get-ApprovedTargetSkuCatalog`.\n2. Display `General Purpose`, `Business Critical`, and `Hyperscale`; require one\n  service type.\n3. Filter and display only the compute\u002Fhardware objectives for that service type;\n  require one objective.\n4. Display only the published vCore values for that objective; require one integer.\n5. Apply the Microsoft Learn documented maximum data size for that vCore\n  automatically; do not present a separate Yes\u002FNo acceptance flow.\n6. Call `Complete-TargetConfiguration`, display the resolved service type,\n  objective, vCore, maximum size, `DatabaseEdition`, and\n  `DatabaseServiceObjective`, then persist them in the sanitized checkpoint.\n\nIf a Phase 4 checkpoint already contains a target configuration, revalidate and\nreuse it rather than asking again. If the checkpoint predates target selection,\nrun the six steps above. Changing an upstream selection resets all downstream\ntracker values.\n\nTreat unresolved target configuration as a blocking interactive state, not a\nsuccessful stopping point. After the export summary, immediately ask for the\nservice type; after each answer, ask for the dependent objective and vCore value\nuntil `Complete-TargetConfiguration` succeeds. Do not end the run, report the\nmigration as complete, or leave imports merely `Pending` while\n`TargetConfiguration` is null unless the user explicitly pauses or cancels.\n\nAfter the modern `sqlcmd` gate passes, default target import authentication to\ninteractive Microsoft Entra ID. Do not ask the user to choose an authentication\nmethod unless they explicitly request the default credential chain. When choices\nmust be shown, present them in this order:\n\n| Target authentication | Command | Recommendation |\n| --- | --- | --- |\n| Interactive Microsoft Entra ID | `sqlcmd -S \u003Cserver> -G -U \u003Cuser>` | **Default**; attended and supports MFA |\n| Microsoft Entra default credential chain (`ActiveDirectoryDefault`) | `sqlcmd -S \u003Cserver> --authentication-method ActiveDirectoryDefault` | Optional explicit override when an Azure CLI, developer credential, or managed identity should be reused |\n\nFor interactive Entra ID, require the user's UPN and never place a password or\ntoken on the command line. Reject SQL login authentication; never invoke a password\nprompt or read credentials from Key Vault, Credential Manager, or a masked control.\nUse the approval-time target preflight as the authentication\u002Fconnectivity check;\ndo not launch a separate preliminary `sqlcmd` probe. After that preflight succeeds,\ncall `New-TargetConnectionStringForApprovedAuthentication` to build the equivalent\nsecretless, `SqlPackage`-supported target connection in memory without another\nbrowser prompt. Do not read an ambient `$targetConnectionString` variable, and do\nnot display or persist the generated value. `SqlPackage` authenticates independently\nwith the selected method; no `sqlcmd` token is handed off. Use\n`InteractiveEntra` unless the user explicitly selects `ActiveDirectoryDefault`.\nSeparate `sqlcmd` and `SqlPackage` processes can each display an interactive\nsign-in prompt.\nThen show the exact Phase 4 approved Azure SQL Database settings for the batch:\n\n- service type (`GeneralPurpose`, `BusinessCritical`, or `Hyperscale`);\n- approved catalog service objective;\n- explicitly selected vCore count;\n- derived SqlPackage database service objective;\n- automatically applied catalog maximum database size in GB.\n\nDo not change SKU settings after the Phase 4 selection is checkpointed. Revalidate\nthe immutable selection against the same approved catalog before displaying it.\nThe single\ngrouped approval must be tabular;\ndo not describe the eligible databases, BACPAC paths, or target settings in a prose\nparagraph. Render these two tables before requesting approval:\n\n| DatabaseName | ExportStatus | BacpacPath |\n| --- | --- | --- |\n| `\u003Ctarget database>` | `Success` | `\u003Cfull local BACPAC path>` |\n\n| TargetServer | Authentication | ServiceType | ServiceObjective | VCore | DatabaseServiceObjective | MaximumSizeGB | Overwrite |\n| --- | --- | --- | --- | ---: | --- | ---: | --- |\n| `\u003Clogical server>` | `\u003Cmethod>` | `\u003CGeneralPurpose \u002F BusinessCritical \u002F Hyperscale>` | `\u003Ccatalog objective>` | `\u003CvCore>` | `\u003Cderived objective>` | `\u003Csize>` | `No` |\n\nAfter authentication, use that same approved identity and validated Go `sqlcmd`\nroute for one injection-safe, read-only `sys.databases` query covering every\nsuccessfully exported target name. Run this preflight before displaying the\napproval tables.\nMark every name observed by this query as `Skipped - database existed before\nimport`, exclude it from the approval table, and retain the observation as\nevidence. If the preflight cannot complete, stop before approval and do not import;\ndo not fall back to diagnostic-text matching.\n\nAfter approval, run exactly one final pre-import preflight and consume its\nscoped `FinalImport` receipt directly in the import loop. There are exactly two\ntarget preflight calls in a successful run: the approval-time call and the final\npre-import call. Never add a third preflight inside a generated import script.\nEach preflight must have a bounded timeout; stop the exact `sqlcmd` process and\npause the migration if it does not return within the configured limit.\nUse a 600-second default for interactive Entra browser\u002FMFA authentication and\n180 seconds for `ActiveDirectoryDefault`. Keep interactive `sqlcmd` attached to\nthe current console and use its `-o` option for structured result capture; do\nnot redirect standard input, output, or error while the attended login runs.\n\nThe final preflight is a point-in-time collision check, not a reservation or\nownership guarantee. With SqlPackage-only target mutation, another actor can\ncreate the name after preflight and before or during import. On success, report\n`TargetStateAfterImport = ImportedSuccessfully`; do not claim that the workflow\ncreated or owns the Azure resource.\n\nAfter the tables, ask one short approval question that states the import is offline\nand authorizes `SqlPackage.exe \u002FAction:Import` against the listed target database\nnames with this residual race disclosed. Do not repeat the table contents in\nprose. After approval, use only\n`SqlPackage.exe \u002FAction:Import` for target mutation and do not call an Azure API.\n\nImport only entries whose `ExportStatus` is `Succeeded`; for each, require its\nselected folder and a file named `\u003Csubfolder-name>.bacpac`. Do not treat a failed\nexport's folder as an import error.\n\nImmediately before each import, validate the BACPAC at the exact manifest path:\n\n- Confirm the expected BACPAC exists and has a plausible nonzero size.\n- Run the BACPAC verification\u002Fchecksum process used by the operator.\n- Do not import when either check fails. Mark that import `Failed` and\n  `AwaitingManualRemediation`, preserve the artifact and diagnostics, and continue\n  only with unaffected databases when `SelectionMode` is `All`.\n\nUse the local import loop in section 4 of\n[references\u002Fcommand-execution-export-import.md](references\u002Fcommand-execution-export-import.md).\n\nUse one `\u002FTargetConnectionString` containing the selected supported authentication\nmethod, server, and database name. Never print that argument. Pass the SKU with\n`\u002Fp:DatabaseEdition`, `\u002Fp:DatabaseServiceObjective`, and\n`\u002Fp:DatabaseMaximumSize`. Do not mix `\u002FTargetConnectionString` with other target\nconnection parameters, including `\u002FTargetDatabaseName`. When constructing the\nper-database connection string with `DbConnectionStringBuilder`, invoke its\n`set_ConnectionString()` and `get_ConnectionString()` methods explicitly;\nPowerShell property assignment can be routed through the builder's dictionary\ninterface and create an invalid nested `ConnectionString` key.\n\nBefore launching the first import, state that live progress will appear in the\nterminal every 30 seconds. Print `Started` immediately for every import, including\nits position in the batch, then poll every 30 seconds and report the database,\nelapsed time, process state, diagnostics-file size and growth, and latest sanitized\ndiagnostic status. Use `Running` when diagnostics changed, `Running - no new\ndiagnostics yet` while the process remains active, and `Possibly stalled` after\nfive minutes without diagnostic activity. The latter is informational and must not\nstop the process before the configured timeout. Import has no growing local output\nfile equivalent to the export BACPAC, so use process state and diagnostics growth\nas movement indicators without claiming a percentage. If an active import reports\n`Incorrect syntax near 'EDITION'` while the target database is already visible as\n`ONLINE`, treat it as a nonterminal SqlPackage diagnostic-masking\u002Fparser message:\nreport it as still running and continue polling until the SqlPackage process exits.\nDo not classify that text alone as a failure, do not restart the import, and do not\nswitch authentication or SKU settings because of it. Do not leave a failed\noperation in `InProgress` or continue polling an exited process. Persist the same\nsanitized lines to the current attempt's\n`\u003Cdatabase-folder>\\attempts\\import\\\u003Cattempt-id>\\status.log`; when an agent or\ncommand runner cannot stream a blocking command's output, it must tail this file\nand relay the latest heartbeat rather than leaving the user without an update.\n\nImport gates:\n\n- Record the `SqlPackage` exit result for every attempted database; only mark an\n  individual import `Succeeded` when its exit code confirms success.\n- Record import duration and any warnings.\n- Stop if the import targets an unexpected server or database.\n\nFor an `All` selection, process only the folders recorded in the manifest. A\ndatabase may be marked `Skipped` only by the successful pre-import\n`sys.databases` observation. For every nonzero SqlPackage exit, record its\nsanitized reason and diagnostics path, mark it `Failed`, and state that the target\nmay have been partially created and requires explicit inspection before retry.\nNever convert an attempted import failure to `Skipped` by matching diagnostic\ntext, regardless of wording or localization. Re-run the same authoritative\ndatabase-existence query after a failed import and record whether the target is\npresent or absent; if that inspection fails, stop the batch rather than continuing\nwith unknown target state. Continue importing the remaining BACPACs when `All` is\nselected only after that inspection succeeds. After all import attempts, display one tabular row per\nattempted database with `DatabaseName`, `Status` (`Success`, `Skipped`, `Fail`, or\n`InProgress`), `ServiceType`, `ServiceObjective`, `VCore`,\n`DatabaseServiceObjective`, and `MaximumSizeGB`. Keep failure\nreasons and diagnostics paths below the table. Invoke validation only for entries\nwhose `ImportStatus` is `Succeeded`.\n\nAfter each database reaches a terminal state, also display the current batch as a\ntable with `DatabaseName`, `Status`, `Elapsed`, and `Detail`. Rows not yet attempted\nmust remain `Pending`; the active row is `InProgress`; completed rows are `Success`,\n`Skipped`, or `Fail`. This intermediate table supplements rather than replaces the\n30-second heartbeat and final import summary.\n\n## Phase 5 — Hand off post-migration validation\n\nFor each manifest entry whose `ImportStatus` is `Succeeded`, invoke\n`validate-post-migration-data` with the existing source connection, target\nconnection details and authentication route, database mapping, source write-freeze\nevidence, approved Azure SQL Database adaptations, and evidence location. Do not\ninvoke validation for skipped exports or imports. Do not request credentials again\nor run data\u002Fschema reconciliation in this migration skill.\n\nStore only the returned validation report status and location in the migration\nmanifest. Do not mark the overall migration complete when any selected database\nreport is `Fail` or `In progress`. This workflow does not delete the source\ndatabase, BACPAC files, diagnostics, manifest, or validation evidence. Cleanup is\na separate, explicitly requested process.\n\n\n\n\n## Operational failures and routing\n\nDo not proactively scan the source for these conditions. Use this table only when\nan export or import command reports the corresponding failure.\n\n| Failure | Action |\n| --- | --- |\n| BACPAC destination file already exists | Do not start or continue that export and do not overwrite, rename, delete, or reuse the file. Mark the item `Failed` and `AwaitingManualRemediation`, save the checkpoint, report the exact path and name collision, and direct the user to preserve\u002Fmove the existing evidence or choose a new export root externally. Resume after `source remediation complete` only when the intended destination no longer exists. |\n| Unsupported `SqlPackage` argument | Stop the batch, mark affected items `AwaitingManualRemediation`, save the checkpoint, and report the argument, installed SqlPackage version, diagnostics, and manual template\u002Fconfiguration correction required. Do not edit or rerun the command inline. Resume failed\u002Fpending databases only after user confirmation. |\n| Unsupported source object or feature | Mark the database `Failed` and `AwaitingManualRemediation`, save the sanitized checkpoint, and provide diagnostics plus external assessment\u002Fchange\u002Ftest\u002Fresume steps. Do not change source code\u002Fschema, restore a substitute database, or invoke remediation inline. Stop for a single database; for `All`, continue only unaffected databases. |\n| Authentication failure during export or import | Mark the database `Failed` and `AwaitingManualRemediation`, save the sanitized checkpoint, and provide identity, permission, network\u002Ffirewall, secure-store or MFA verification steps appropriate to the selected method. Do not retry, refresh, or switch authentication inline. |\n| Target database exists before import | Observe it with the injection-safe, read-only `sys.databases` preflight through validated Go `sqlcmd`, mark it `Skipped - database existed before import`, and exclude it from import approval\u002Fexecution. A SqlPackage failure never proves this condition. |\n| Target may have been created by a failed import | Keep the attempted import `Failed`, preserve diagnostics, and require explicit target inspection before retry. Do not relabel it `Skipped` merely because a database now exists. |\n| `Incorrect syntax near 'EDITION'` appears while import is still running and target database is `ONLINE` | Treat as nonterminal SqlPackage diagnostic-masking\u002Fparser output. Continue polling the active process and use only process exit code plus authoritative target inspection after exit for terminal classification. |\n| Import timeout or throttling | Mark the import `Failed` and `AwaitingManualRemediation`, save the checkpoint, and instruct the user to address target sizing, throttling, or connectivity externally. Retry the same verified BACPAC only after `target remediation complete`. |\n| Missing target user or permission | Route identity remediation to the target security workflow. |\n| Target capacity is insufficient | Stop and route to the sizing\u002Fplanning skill before retrying import. |\n\n## Manual-remediation resume gate\n\nPersist a sanitized checkpoint after every terminal export\u002Fimport result. It may\ncontain nonsecret mappings, statuses, timestamps, diagnostics paths, failure\ncategories, and manual next actions, but never connection strings, passwords,\ntokens, or command lines containing them. On resume, treat the checkpoint only as\na cursor: revalidate the source\u002Ftarget identity, selected database mapping, local\nBACPAC state, and target preflight from authoritative sources.\n\nRetry an `AwaitingManualRemediation` item only after the user explicitly states\n`authentication remediation complete`, `source remediation complete`, or `target\nremediation complete` as applicable. Display the failed database, prior sanitized\nreason, and intended retry before proceeding. Preserve successful exports\u002Fimports\nand do not repeat them. If the user instead requests a fresh execution, rebuild\nthe manifest after discovery in a new export root. Never overwrite an existing\ncheckpoint, BACPAC, diagnostics file, or status file during a fresh execution;\nretain the old export root as immutable evidence.\n\nEvery export or import invocation, including an approved retry, must create a new\nUTC-and-GUID attempt directory. Record each attempt ID, timestamps, result, and\nits diagnostics, console, and status paths in the sanitized checkpoint. Never\nreuse or truncate an earlier attempt's evidence files.\n\n## Result format\n\nAfter planning, report:\n\n> **Migration plan — `\u003Csource database or all selectable user databases>`**\n> **Azure SQL Database** via **BACPAC** · downtime **offline**\n\n| | Decision |\n| --- | --- |\n| Target | Azure SQL Database — `\u003Clogical server name>` |\n| Method | BACPAC export\u002Fimport with `SqlPackage` |\n| Import execution | `SqlPackage.exe \u002FAction:Import` only |\n| Completion | `\u003CComplete \u002F Failed \u002F In progress>` |\n\nInclude:\n\n- Selected database scope and resolved export root.\n- One local BACPAC path per database.\n- Per-database export\u002Fimport operation status.\n- Last status-update time, sanitized failure reason, and diagnostics location for\n  every failed or long-running operation.\n- Per-database validation handoff\u002Freport status and report location.\n\nFor a failed operation, state the failed phase, error category, evidence\nlocation, and the remediation required. Never present a partial migration as\ncomplete.\n\n## References\n\n- [Local BACPAC command execution — manifest setup](references\u002Fcommand-execution-setup.md)\n- [Local BACPAC command execution — export and import](references\u002Fcommand-execution-export-import.md)\n- [SqlPackage export action](https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Fsql\u002Ftools\u002Fsqlpackage\u002Fsqlpackage-export)\n- [SqlPackage import action](https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Fsql\u002Ftools\u002Fsqlpackage\u002Fsqlpackage-import)\n- [Azure SQL Database migration overview](https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Fdata-migration\u002Fsql-server\u002Fdatabase\u002Foverview)\n- [Azure SQL Database elastic jobs](https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Fazure\u002Fazure-sql\u002Fdatabase\u002Felastic-jobs-overview)\n",{"name":5,"description":7},"main","d2b16500cb10bf44117390e08717ca731032d423","corenote\u002Fmigration\u002Fskills\u002Fsql-bacpac-to-azure-sql-db-migration",0,"https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fsqlcon-eu-2026-demos","https:\u002F\u002Fgithub.com\u002Fmicrosoft","https:\u002F\u002Fopensource.microsoft.com","Open source projects and samples from Microsoft","2026-11-01 23:31:12",{"source":8,"skillName":5,"commands":50,"prompt":60,"zipAvailable":61},[51,54,57],{"tool":52,"command":53},"npx","npx skills add microsoft\u002Fsqlcon-eu-2026-demos --skill sql-bacpac-to-azure-sql-db-migration",{"tool":55,"command":56},"bunx","bunx skills add microsoft\u002Fsqlcon-eu-2026-demos --skill sql-bacpac-to-azure-sql-db-migration",{"tool":58,"command":59},"pnpm","pnpm dlx skills add microsoft\u002Fsqlcon-eu-2026-demos --skill sql-bacpac-to-azure-sql-db-migration","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fsqlcon-eu-2026-demos\u002Fblob\u002Fmain\u002Fcorenote\u002Fmigration\u002Fskills\u002Fsql-bacpac-to-azure-sql-db-migration\u002FSKILL.md",true,[63,69,75,79,85,89,93,97,101,105,109],{"path":64,"resourceType":65,"resourceTypeDesc":66,"sizeBytes":67,"sourceHtmlUrl":68},"assets\u002Fadr_template.md","ASSET","资产文件",6950,"https:\u002F\u002Fgithub.com\u002Fshengdabai\u002FTony-Claude-Code-Skills\u002Fblob\u002Fmain\u002Fskills\u002F_collections\u002Frghamilton3-dotfiles\u002Fdot_claude\u002Fskills\u002Fiot-system-architect\u002Fassets\u002Fadr_template.md",{"path":70,"resourceType":71,"resourceTypeDesc":72,"sizeBytes":73,"sourceHtmlUrl":74},"references\u002Fargus_architecture.md","REFERENCE","参考文档",13366,"https:\u002F\u002Fgithub.com\u002Fshengdabai\u002FTony-Claude-Code-Skills\u002Fblob\u002Fmain\u002Fskills\u002F_collections\u002Frghamilton3-dotfiles\u002Fdot_claude\u002Fskills\u002Fiot-system-architect\u002Freferences\u002Fargus_architecture.md",{"path":76,"resourceType":71,"resourceTypeDesc":72,"sizeBytes":77,"sourceHtmlUrl":78},"references\u002Fazure-sql-db-vcore-sku-limits.json",6031,"https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fsqlcon-eu-2026-demos\u002Fblob\u002Fmain\u002Fcorenote\u002Fmigration\u002Fskills\u002Fsql-bacpac-to-azure-sql-db-migration\u002Freferences\u002Fazure-sql-db-vcore-sku-limits.json",{"path":80,"resourceType":81,"resourceTypeDesc":82,"sizeBytes":83,"sourceHtmlUrl":84},"references\u002Fbacpac-checkpoint.ps1","SCRIPT","脚本",7004,"https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fsqlcon-eu-2026-demos\u002Fblob\u002Fmain\u002Fcorenote\u002Fmigration\u002Fskills\u002Fsql-bacpac-to-azure-sql-db-migration\u002Freferences\u002Fbacpac-checkpoint.ps1",{"path":86,"resourceType":81,"resourceTypeDesc":82,"sizeBytes":87,"sourceHtmlUrl":88},"references\u002Fbacpac-target-preflight.ps1",14118,"https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fsqlcon-eu-2026-demos\u002Fblob\u002Fmain\u002Fcorenote\u002Fmigration\u002Fskills\u002Fsql-bacpac-to-azure-sql-db-migration\u002Freferences\u002Fbacpac-target-preflight.ps1",{"path":90,"resourceType":71,"resourceTypeDesc":72,"sizeBytes":91,"sourceHtmlUrl":92},"references\u002Fcommand-execution-export-import.md",42554,"https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fsqlcon-eu-2026-demos\u002Fblob\u002Fmain\u002Fcorenote\u002Fmigration\u002Fskills\u002Fsql-bacpac-to-azure-sql-db-migration\u002Freferences\u002Fcommand-execution-export-import.md",{"path":94,"resourceType":71,"resourceTypeDesc":72,"sizeBytes":95,"sourceHtmlUrl":96},"references\u002Fcommand-execution-setup.md",47802,"https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fsqlcon-eu-2026-demos\u002Fblob\u002Fmain\u002Fcorenote\u002Fmigration\u002Fskills\u002Fsql-bacpac-to-azure-sql-db-migration\u002Freferences\u002Fcommand-execution-setup.md",{"path":98,"resourceType":71,"resourceTypeDesc":72,"sizeBytes":99,"sourceHtmlUrl":100},"references\u002Fdesign_patterns.md",16732,"https:\u002F\u002Fgithub.com\u002Fshengdabai\u002FTony-Claude-Code-Skills\u002Fblob\u002Fmain\u002Fskills\u002F_collections\u002Frghamilton3-dotfiles\u002Fdot_claude\u002Fskills\u002Fiot-system-architect\u002Freferences\u002Fdesign_patterns.md",{"path":102,"resourceType":71,"resourceTypeDesc":72,"sizeBytes":103,"sourceHtmlUrl":104},"references\u002Ftarget-database-preflight.sql",647,"https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fsqlcon-eu-2026-demos\u002Fblob\u002Fmain\u002Fcorenote\u002Fmigration\u002Fskills\u002Fsql-bacpac-to-azure-sql-db-migration\u002Freferences\u002Ftarget-database-preflight.sql",{"path":106,"resourceType":81,"resourceTypeDesc":82,"sizeBytes":107,"sourceHtmlUrl":108},"references\u002Ftarget-sku-validation.ps1",15047,"https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fsqlcon-eu-2026-demos\u002Fblob\u002Fmain\u002Fcorenote\u002Fmigration\u002Fskills\u002Fsql-bacpac-to-azure-sql-db-migration\u002Freferences\u002Ftarget-sku-validation.ps1",{"path":110,"resourceType":71,"resourceTypeDesc":72,"sizeBytes":111,"sourceHtmlUrl":112},"skill-contract.yml",4994,"https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fsqlcon-eu-2026-demos\u002Fblob\u002Fmain\u002Fcorenote\u002Fmigration\u002Fskills\u002Fsql-bacpac-to-azure-sql-db-migration\u002Fskill-contract.yml",[114,124,135,145,156,165],{"skillNo":115,"name":116,"nameZh":6,"description":117,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":118,"categories":119,"keywords":121,"publishedAt":35},"01M69AGTKN8F66BXE45KQ3MDMX","keynote-demo-app","Build or extend a SQLCon EU 2026 keynote demo application so it looks and behaves like the other keynote apps. Use when creating a new demo view (invoices, operations, orders), restyling an existing app to match, adding a deliberate database anti-pattern for a troubleshooting demo, or wiring an app to Azure SQL or Fabric SQL with managed identity. Triggers include \"make this look like the keynote app\", \"build a demo view\", \"add the slow query demo\", \"match Caldova styling\".","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fsqlcon-eu-2026-demos\u002Fblob\u002Fmain\u002Fshared\u002FSKILL.md",[120],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[122,123,29,30],"keynote demo app","identity",{"skillNo":125,"name":126,"nameZh":6,"description":127,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":128,"categories":129,"keywords":132,"publishedAt":35},"01M69AGRAPA61AN2A2N2RF6NCJ","validate-post-migration-data","Validate data and schema after migrating SQL Server to Azure SQL Database, Azure SQL Managed Instance, or SQL Server on Azure VM. Use after migration, restore, BACPAC import, LRS completion, cutover validation, row-count comparison, or source-target reconciliation.","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fsqlcon-eu-2026-demos\u002Fblob\u002Fmain\u002Fcorenote\u002Fmigration\u002Fskills\u002Fvalidate-post-migration-data\u002FSKILL.md",[130,131],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},{"categoryNo":24,"categoryName":25,"taxonomyType":26,"taxonomyTypeDesc":27},[133,134,34,30,33,29],"validate","validate post migration data",{"skillNo":136,"name":137,"nameZh":6,"description":138,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":139,"categories":140,"keywords":143,"publishedAt":35},"01M69AGNG0HQK50EW1CNEFTR1N","sql-server-to-sql-mi-lrs-migration","Migrate SQL Server databases to Azure SQL Managed Instance with Log Replay Service (LRS) using managed identity for Blob access. Use for SQL Server to SQL MI, LRS, log replay, continuous restore, or low-downtime backup-chain migrations.","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fsqlcon-eu-2026-demos\u002Fblob\u002Fmain\u002Fcorenote\u002Fmigration\u002Fskills\u002Fsql-server-to-sql-mi-lrs-migration\u002FSKILL.md",[141,142],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},{"categoryNo":24,"categoryName":25,"taxonomyType":26,"taxonomyTypeDesc":27},[144,34,30,123,29],"sql server to sql mi lrs migration",{"skillNo":146,"name":147,"nameZh":6,"description":148,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":149,"categories":150,"keywords":153,"publishedAt":35},"01M69AGFTTX13PCFKM3YVHKGDV","sql-backup-restore-to-azure-sql-vm-migration","Migrate a SQL Server 2008 through SQL Server 2025 source to a SQL Server 2025 target wherever it runs, including SQL Server on Azure VM, by creating a local full backup, uploading it to Azure Blob Storage with AzCopy and the operator's Microsoft Entra identity, and restoring from URL with the target SQL Server managed identity. Sources newer than SQL Server 2025 are blocked because target restore compatibility has not been validated. Use for SQL-to-SQL, local-backup, AzCopy, managed-identity, Azure SQL VM, or SQL Server 2025 restore migrations.","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fsqlcon-eu-2026-demos\u002Fblob\u002Fmain\u002Fcorenote\u002Fmigration\u002Fskills\u002Fsql-backup-restore-to-azure-sql-vm-migration\u002FSKILL.md",[151,152],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},{"categoryNo":24,"categoryName":25,"taxonomyType":26,"taxonomyTypeDesc":27},[154,30,29,34,123,9,155],"sql backup restore to azure sql vm migration","storage",{"skillNo":157,"name":158,"nameZh":6,"description":159,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":160,"categories":161,"keywords":163,"publishedAt":35},"01M69AGDP6XXM1YW7X1VV1QKE5","run-migration-assessment","Route generic or unresolved SQL Server assessment, readiness, compatibility, cost, and SKU recommendation requests. Use for bare server names or whenever Azure versus local\u002Fon-premises is not explicit; ask which source applies, then delegate.","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fsqlcon-eu-2026-demos\u002Fblob\u002Fmain\u002Fcorenote\u002Fmigration\u002Fskills\u002Frun-migration-assessment\u002FSKILL.md",[162],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[164,34,30],"run migration assessment",{"skillNo":166,"name":167,"nameZh":6,"description":168,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":169,"categories":170,"keywords":171,"publishedAt":35},"01M69AGBJN9ESKFXD8YXB7J20W","recommend-sku-sizing","Collect performance data and calculate Azure SQL SKU recommendations only after the source is explicitly confirmed local\u002Fon-premises. Never use for a generic SKU request or bare server name; route unresolved requests through run-migration-assessment.","https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fsqlcon-eu-2026-demos\u002Fblob\u002Fmain\u002Fcorenote\u002Fmigration\u002Fskills\u002Frecommend-sku-sizing\u002FSKILL.md",[],[172,173,30,29],"recommend sku sizing","performance"]