agentleFS
MicrosoftKnown publishermicrosoft/sqlsimtools/.github/skills/sqlsimtools-release/SKILL.md

sqlsimtools-release

Skill42 starsChanged 32 days agoMITView on GitHub

Build, package, validate, and publish sqlsimtools release assets. USE WHEN: running a release candidate dry run, cutting a version, tagging a release, publishing a draft GitHub Release, verifying archive contents, diagnosing a failed release workflow, or changing any publish script. DO NOT USE FOR: running sqlsim or sqlpysim workloads (use the sqlsim or sqlpysim skills), or day-to-day source edits.

  • Commits and pushes

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 6 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    When adding or renaming any staged file, update every affected script in the same change, then add the path to …

    line 117
  2. Very specific:

    Only the tag path has contents: write.

    line 139
  3. Distinctive:

    Always dry run before tagging.

    line 160
  4. Distinctive:

    Use semantic versioning: patch for a fix, minor for a compatible feature, major for a breaking CLI or output change.

    line 189
  5. Distinctive:

    Do not move a published tag, replace assets on an existing release, or delete older releases.

    line 202
  6. Distinctive:

    Never tag without a passing dry run on the same commit.

    line 214
  7. Distinctive:

    Never edit one publish script without checking the other three.

    line 215
  8. Very specific:

    Every archive ships LICENSE and THIRD-PARTY-NOTICES.md.

    line 216
  9. Distinctive:

    Treat a validation failure as a packaging defect, not a flaky run.

    line 219

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

sqlsimtools release

Developer skill for producing and publishing release assets. This is not product documentation and does not ship in release archives.

Publishing requires write access to the repository. Contributors without it can still use this skill to understand what an archive must contain and to verify a downloaded release, but the tagging and publishing steps are maintainer-only. To build and test from source instead, use the sqlsimtools-build skill.

Release assets

Five public assets. The sqlsim and sqlpysim archives are independent; the sqlsimtools archive bundles sqlsim Studio with the Windows x64 sqlsim archive and the sqlpysim archive, unchanged.

AssetProduced byRunner
sqlsim-<ver>-win-x64.zipsqlsim/publish.ps1windows-latest
sqlsim-<ver>-win-arm64.zipsame run, split per architecturewindows-latest
sqlsim-<ver>-linux-x64.tar.gzsqlsim/publish-linux.shubuntu-24.04
sqlpysim-<ver>-src.zipsqlpysim/publish.ps1windows-latest
sqlsimtools-<ver>-win-x64.ziprelease/build-sqlsimtools-bundle.ps1 from the sqlsim x64 and sqlpysim archives plus sqlsimstudio/build.ps1 outputwindows-latest

release/stage-release-assets.ps1 attaches these five versioned packages and nothing else. The README links to releases/latest, the newest published release page, so it never needs a version edit. The script fails if any package is missing.

The workflow also builds and validates sqlsimstudio-<ver>-linux-x64.tar.gz as an Actions artifact for invited Private Preview testing. It is not attached to the public GitHub Release.

<ver> has no v prefix. A tag v1.0.0 produces 1.0.0.

Required archive contents

The sqlsim and sqlpysim archives are project overlays. They extract at a consumer project root and must not nest their payload any deeper.

.github/agents/sqlsimtools-workload-builder.agent.md
.github/skills/<tool>/SKILL.md
<tool>/AGENTS.md
<tool>/LICENSE
<tool>/THIRD-PARTY-NOTICES.md
<tool>/<tool>.md
<tool>/examples/

Plus the payload: sqlsim/x64/sqlsim.exe and sqlsim/ARM64/sqlsim.exe on Windows, sqlsim/x64/sqlsim on Linux, or sqlpysim/sqlpysim.py with requirements.txt and setup.py.

The sqlsimtools archive is both overlays plus Studio in its own folder. Studio runs the bundled engines from ../sqlsim/x64 and ../sqlpysim, so the bundle carries no second copy of either:

README.md
.github/skills/{sqlsim,sqlpysim,sqlsimstudio}/SKILL.md
.github/agents/sqlsimtools-workload-builder.agent.md
sqlsim/...              the complete sqlsim win-x64 archive
sqlpysim/...            the complete sqlpysim archive
sqlsimstudio/sqlsimstudio.exe
sqlsimstudio/sqlsimstudio.md
sqlsimstudio/sqlsimstudio.html
sqlsimstudio/CONFIGURATION.md
sqlsimstudio/sqlsimstudio-cleanup.ps1
sqlsimstudio/examples/
sqlsimstudio/LICENSE
sqlsimstudio/THIRD-PARTY-NOTICES.md

The bundle script refuses differing overlapping files, an ARM64 sqlsim payload, and missing entry points. The build-sqlsimtools job then validates the bundle with validate-package-layout.ps1 for both tools, the Studio startup smoke test, and the Studio offline suite, which proves Studio finds the bundled sqlsim. The Private Preview Linux Studio artifact remains a standalone Studio build with its own engine copies:

sqlsimstudio
sqlsim/sqlsim
sqlpysim/sqlpysim.py
sqlpysim/requirements.txt
.github/skills/sqlsimstudio/SKILL.md
sqlsimstudio.md
CONFIGURATION.md
examples/
LICENSE
THIRD-PARTY-NOTICES.md

Studio also ships native libraries beside the executable. PublishSingleFile bundles managed assemblies only; the operating system loader requires native libraries on disk. Do not treat icudt72.dll, icuin72.dll, icuuc72.dll, or Microsoft.Data.SqlClient.SNI.dll as packaging defects. App-local ICU is pinned deliberately so string comparison and sorting stay identical across machines.

Parity rule

Four publish paths stage the same logical layout and drift independently:

  • sqlsim/publish.ps1
  • sqlsim/publish-linux.sh
  • sqlpysim/publish.ps1
  • sqlsimstudio/build.ps1 and sqlsimstudio/build-linux.sh
  • release/build-sqlsimtools-bundle.ps1

When adding or renaming any staged file, update every affected script in the same change, then add the path to release/validate-package-layout.ps1 or to the sqlsimtools and Linux Studio checks in .github/workflows/release.yml. A file that only one platform stages is a defect that reaches users as a missing license, a missing agent file, or a failed release job.

The sqlsim and sqlpysim publishers also install into existing projects, so they must never delete project files or silently replace one that differs. sqlsim/publish.ps1 and sqlpysim/publish.ps1 share release/OverlayInstall.psm1 for this; sqlsim/publish-linux.sh implements the same rules. Keep release/test-overlay-install.ps1 and release/test-overlay-install-linux.sh passing; sqlsim CI runs both.

Workflow triggers

.github/workflows/release.yml runs two ways.

TriggerResult
workflow_dispatch with a version inputBuilds and validates every asset, uploads GitHub Actions artifacts only. No tag, no Release.
Push a tag matching v*.*.*Builds everything and creates a draft GitHub Release for review.

Only the tag path has contents: write. Manual runs cannot publish.

The workflow does not modify in-tree version constants. Its first job, verify-version, runs release/check-release-version.ps1 and fails the run unless the tag or RC label (ignoring a v prefix and any -rc suffix) matches the version in every one of these files:

  • sqlsim/include/version.h (SQLSIM_VERSION)
  • sqlsim/CMakeLists.txt (project(sqlsim VERSION ...))
  • sqlpysim/sqlpysim.py (__version__)
  • sqlpysim/setup.py (version=)
  • sqlsimstudio/server/SqlsimStudio.csproj (<Version>)

Bump all five in the release PR, then confirm locally before tagging:

./release/check-release-version.ps1 -Version v1.0.1

Release candidate dry run

Always dry run before tagging.

gh workflow run release.yml --ref main -f version=1.0.0-rc1
gh run list --workflow=release.yml --limit 1
gh run watch <run-id>

Download and inspect the artifacts:

gh run download <run-id> --dir ./rc-artifacts

For each archive, extract to an empty directory and confirm:

  • The overlay paths above exist, with no extra wrapper directory.
  • sqlsim --version and sqlsim -h run from the extracted copy.
  • LICENSE and THIRD-PARTY-NOTICES.md are present in every archive.
  • The workload-builder agent is present in both sqlsim and sqlpysim archives.

The workflow already extracts each archive and runs release/validate-package-layout.ps1 against it, runs the Studio smoke test and offline suite against the sqlsimtools bundle, and smoke tests the Private Preview Linux Studio build. A green run means layout and startup passed; it does not mean the tools were exercised against a live SQL target.

Publish a version

Use semantic versioning: patch for a fix, minor for a compatible feature, major for a breaking CLI or output change.

git checkout main
git pull
git tag v1.0.0
git push origin v1.0.0

The tag starts the workflow and creates a draft Release with generated notes. Review the draft and its assets, then publish it in the Releases UI.

Do not move a published tag, replace assets on an existing release, or delete older releases. Ship a new patch version instead.

Local builds

release/publish.ps1 -PublishAll runs each tool's publish script in sequence and stages output under release/sqlsimtools/1.0/. That directory is ignored by Git, is not the release pipeline, and can be stale. Never use it to judge what an archive contains. Use artifacts from a dry run instead.

Rules

  1. Never tag without a passing dry run on the same commit.
  2. Never edit one publish script without checking the other three.
  3. Every archive ships LICENSE and THIRD-PARTY-NOTICES.md.
  4. Parse JSON with a structured parser. Do not append 2>&1 or pipe tool output through text filters when capturing results.
  5. Treat a validation failure as a packaging defect, not a flaky run.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.