usage-monitor-release-beta
Skill19 starsChanged 50 days agoMITView on GitHub
Beta channel of usage-monitor — publish a vX.Y.Z-beta.N prerelease for opted-in users, fix it with the next beta, promote the series to stable, or withdraw/restore a beta. Use for any beta release task; plain stable releases use usage-monitor-release.
- Commits and pushes
In this repository13 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- orchestrateSkill
- usage-monitor-commit-pushSkill
- usage-monitor-designSkill
- usage-monitor-nsis-installerSkill
- usage-monitor-releaseSkill
- usage-monitor-release-videoSkill
- usage-monitor-social-videoSkill
- usage-monitor-test-reportSkill
- usage-monitor-token-cleanupSkill
- usage-monitor-visual-optionsSkill
- usage-monitor-workflowSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
19 rules. About 0 are common, 9 are distinctive, and 10 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈9
Distinctive · estimate
This author's own choices. Worth a look.
10
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
Run the whole life of a beta for the users who opted into the beta channel (issue #355).
- Very specific: line 48
Read AGENTS.md, build.gradle.kts and .github/workflows/release-linux.yml first.
- Distinctive: line 58
If the request is ambiguous and there is no open series, ask which type.
- Very specific: line 60
Verify that main is clean and matches origin/main — the tag must point to a commit already on the remote main, or verify-version refuses it:
- Distinctive: line 71
Run the local verification that must pass before the tag exists:
- Distinctive: line 80
Do not package locally as a gate.
- Very specific: line 90
every artifact family is there, each with a sha256: digest (auto-update refuses assets without it): UsageMonitor-Setup-X.Y.Z-beta.N.exe.
- Very specific: line 113
default — the current origin/main (the beta code plus whatever was merged after it).
- Very specific: line 118
Run the stable flow of usage-monitor-release from its step 4 (clean main = origin/main, allTests, the #342 window-host question, annotated …
- Distinctive: line 152
If the maintainer really wants it, tell them the command and let them run it.
- Distinctive: line 160
If a newer beta or the stable of the series already exists, restoring an older beta changes nothing for anyone — say so instead of …
- Very specific: line 165
Never publish a stable version from this skill except through promote, which runs the stable flow of usage-monitor-release with the version …
- Very specific: line 167
Never turn a beta release into a stable or Latest one by editing it.
- Very specific: line 169
Never delete a release or a tag.
- Very specific: line 170
Never tag a suffix other than -beta.N.
- Very specific: line 171
Never tag a beta whose X.Y.Z is already released as stable.
- Distinctive: line 172
Do not release from a dirty tree unless the user explicitly wants that risk.
- Distinctive: line 173
Do not ask the user to confirm the tag or the push.
- Distinctive: line 174
Prefer current code and workflow files over older notes when they disagree.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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.
Files
Usage Monitor Beta Release
Run the whole life of a beta for the users who opted into the beta channel (issue #355). Publishing
is the stable release flow (usage-monitor-release) with a
different tag and a different promise: a beta never becomes latest, so nobody outside the
channel receives it.
Operations (the argument picks one)
| Argument | What it does | Section |
patch / minor / major | Publish beta.1 of a new series for the next stable of that type | Workflow |
next | Publish beta.N+1 of the open series (the way to fix a bad beta) | Workflow |
promote | Publish the stable vX.Y.Z of the open series — the beta "goes to main/master" | Promote |
withdraw [tag] | Hide a beta: nobody new is offered it (reversible) | Withdraw |
restore [tag] | Undo a withdraw | Restore |
There is no "move the beta to main": the tag already points to a commit on main —
verify-version refuses anything else. The code is merged before the beta exists; what "going to
stable" means is publishing a new, stable build of that code (promote). There is also no
rollback for testers who already installed a beta: the app never downgrades. They leave the beta
when a higher version reaches them — the next beta or the stable.
How the channel works (read before tagging)
- Tag format:
vX.Y.Z-beta.N,N ≥ 1. Nothing else is accepted —verify-versionrejects-rc.1,-betawithout a number and any other suffix, and the Linux updater refuses them too. - The workflow sets
prerelease: trueandmake_latest: falsefor-beta.Ntags. Clients outside the channel read/releases/latest, which the GitHub API never answers with a prerelease — that is the whole protection, and it also covers app versions older than the channel. - Clients inside the channel list
/releases?per_page=20and take the highest non-draft version by SemVer precedence:42.0.0-beta.1 < 42.0.0-beta.2 < 42.0.0. The stable42.0.0therefore reaches every beta tester of the42.0.0series. Turning the channel off never downgrades. X.Y.Zis the next stable version the beta previews. Never tag a beta for a number that is already released as stable:41.0.0-beta.1afterv41.0.0is lower than the running stable and is never offered to anyone.- Packaging: Exe/Dmg carry only
X.Y.Z(jpackage rejects the suffix), Deb/Rpm carryX.Y.Z~beta.N;CURRENT_APP_VERSION, the NSIS installer name, the tarball and the release assets keep the fullX.Y.Z-beta.N.
Workflow
- Read AGENTS.md, build.gradle.kts and .github/workflows/release-linux.yml first.
- Invoking this skill already means "cut and publish the beta". Do not ask for confirmation to tag or push — go through the whole flow and report the result. The only question allowed is the one in step 6 (window hosts).
- Pick the operation from the request (see the table above).
promote,withdrawandrestorehave their own sections below; this workflow is for publishing a beta:patch/minor/major— start a new series for the next stable of that type:N = 1.next(or no type while a series is open) — continue the open series:N + 1.- If the request is ambiguous and there is no open series, ask which type.
- An open series is
vX.Y.Z-beta.*with no stablevX.Y.Zyet.
- Verify that
mainis clean and matchesorigin/main— the tag must point to a commit already on the remotemain, orverify-versionrefuses it:git fetch origin --tags && git status --short --branchgit log --oneline -5
- Compute the version. There is no version bump commit — the version comes from the tag.
- Last stable:
git describe --tags --abbrev=0 --match "v[0-9]*" --exclude "*-beta*". - Betas of the series:
git tag --list "vX.Y.Z-beta.*", highestNby numeric sort (git -c versionsort.suffix=- tag --sort=-version:refname --list "vX.Y.Z-beta.*" | head -n 1). - New series:
X.Y.Z= last stable bumped by the type; refuse ifvX.Y.Zalready exists. - Continue: same
X.Y.Z,N + 1; refuse if the stablevX.Y.Zalready exists (the series is closed — start a new one).
- Last stable:
- Run the local verification that must pass before the tag exists:
gradlew.bat allTests— mandatory. A beta is still a public tag; a broken suite discovered after the push is a burned tag.- Window hosts changed? Stop before the tag (#342), exactly as in the stable skill: run
git diff --name-only $(git describe --tags --abbrev=0)..HEADand look forHudWindow.kt,DesktopWindowFrame.kt,Main.ktor any*WindowHost*.kt. If one changed, list the rows ofdocs/hud-notch.md› "Fora do alcance dos testes" the change touches and ask the maintainer whether the build was opened on Linux (X11). A beta is a good place to find such bugs, but the testers still run it as their daily app. - Do not package locally as a gate; CI owns packaging for the three OSes.
- Publish:
- annotated tag with the temporary agent git identity:
git -c user.name=claude -c user.email=claude@anthropic.com tag -a vX.Y.Z-beta.N -m "vX.Y.Z-beta.N" - push only the tag:
git push origin vX.Y.Z-beta.N.
- annotated tag with the temporary agent git identity:
- Watch
Release Desktop Packagesand report the outcome. Then confirm on the published release:- it is marked Pre-release:
gh release view vX.Y.Z-beta.N --json isPrerelease→true(ghhas noisLatestfield); - it is not Latest —
/releases/lateststill points to the last stable:gh api repos/edilsonvilarinho/usage-monitor/releases/latest --jq .tag_name; - every artifact family is there, each with a
sha256:digest (auto-update refuses assets without it):UsageMonitor-Setup-X.Y.Z-beta.N.exe;.deb,.rpmandusage-monitor_X.Y.Z-beta.N_linux_x64.tar.gz;usage-monitor_X.Y.Z-beta.N_macos_arm64.dmgandusage-monitor_X.Y.Z-beta.N_macos_x64.dmg.gh release view vX.Y.Z-beta.N --json assets --jq '.assets[] | [.name, .digest] | @tsv'A beta missing an artifact family, or published as Latest, is a failed release and must be reported as such. If it went out as Latest, edit it back withgh release edit vX.Y.Z-beta.N --prerelease --latest=falseand report it — every user outside the channel was being offered the beta.
- it is marked Pre-release:
- The release notes come from the workflow's
Generate release notesstep: a beta diffs against the previous tag (beta or stable); the stable release later diffs against the previous stable, so it lists the whole series. Do not write them by hand.
Promote — the open series becomes the stable release
The stable is a new tag and a new build, never the beta release edited into a stable one. The beta
binaries carry X.Y.Z-beta.N in CURRENT_APP_VERSION, in the installer name and in the receipt; an
edited release would hand every user a build that calls itself beta. Never run
gh release edit ... --prerelease=false or --latest on a beta.
- Find the open series:
git -c versionsort.suffix=- tag --sort=-version:refname --list "v*-beta.*" | head -n 1givesvX.Y.Z-beta.N. Refuse ifvX.Y.Zalready exists (nothing to promote) or if there is no beta tag. - Decide what becomes stable:
- default — the current
origin/main(the beta code plus whatever was merged after it). Say in the report which commits are new since the last beta:git log --oneline vX.Y.Z-beta.N..origin/main. promote exact— the very commit the testers ran: tagvX.Y.ZonvX.Y.Z-beta.N^{commit}.verify-versionaccepts it, because that commit is onmain.
- default — the current
- Run the stable flow of usage-monitor-release from its step 4
(clean
main=origin/main,allTests, the #342 window-host question, annotated tag, push only the tag, watch the workflow, check every artifact) with the version fixed toX.Y.Z— do not bump it by type. Forpromote exact, create the tag on the beta commit:git -c user.name=claude -c user.email=claude@anthropic.com tag -a vX.Y.Z vX.Y.Z-beta.N^{commit} -m "vX.Y.Z". - Confirm:
gh release view vX.Y.Z --json isPrerelease→false, and/releases/latest→vX.Y.Z. The stable notes diff against the previous stable, so they list the whole beta series. - Result: users outside the channel receive
X.Y.Zas a normal update; testers onX.Y.Z-beta.Nreceive it too (beta.N < X.Y.Z) and see its "Novidades". The series is closed — the next beta starts a new one.
Withdraw — a bad beta stops being offered
Withdrawing hides the release; it does not delete anything and it is reversible (restore).
The client ignores drafts, and anonymous API calls — the ones the app makes — do not see them.
- Target: the tag in the request, or the highest
v*-beta.*. Refuse a stable tag: withdrawing a stable is not this skill's job. gh release edit vX.Y.Z-beta.N --repo edilsonvilarinho/usage-monitor --draft=true- Confirm it is hidden from what the app reads:
gh api "repos/edilsonvilarinho/usage-monitor/releases?per_page=20" --jq '.[] | select(.draft == false) | .tag_name'must not list it (this call is authenticated and sees drafts only withdraft: true), and/releases/latestis unchanged. - Report what it does not undo, every time:
- testers who already installed that beta stay on it — there is no downgrade. They move when a higher version reaches them;
- an auto-update already downloaded but not yet applied is still applied on the next exit.
- What to do next — say it in the report:
- fix forward (the normal path): merge the fix and run this skill with
next—vX.Y.Z-beta.N+1is offered to everyone in the channel, including those stuck on the bad one; - abandon the series:
promotea stable that does not include the bad change, or start a new series later. Never reuse a withdrawn tag name — a client may already know it.
- fix forward (the normal path): merge the fix and run this skill with
- The tag stays on the remote. Deleting the release or the tag (
gh release delete --cleanup-tag) is permanent and not part of this skill: it gains nothing over the draft and loses the record. If the maintainer really wants it, tell them the command and let them run it.
Restore — undo a withdraw
gh release edit vX.Y.Z-beta.N --repo edilsonvilarinho/usage-monitor --draft=false --prerelease --latest=false— the three flags together, so it comes back exactly as a beta and never as Latest.- Confirm
isPrerelease: true,/releases/latestunchanged, and that the channel sees it again (the non-draft listing above lists it). - If a newer beta or the stable of the series already exists, restoring an older beta changes nothing for anyone — say so instead of restoring.
Guardrails
- Never publish a stable version from this skill except through
promote, which runs the stable flow ofusage-monitor-releasewith the version fixed to the open series. - Never turn a beta release into a stable or Latest one by editing it;
restorealways passes--prerelease --latest=false. - Never delete a release or a tag;
withdrawhides it as a draft. - Never tag a suffix other than
-beta.N. - Never tag a beta whose
X.Y.Zis already released as stable. - Do not release from a dirty tree unless the user explicitly wants that risk.
- Do not ask the user to confirm the tag or the push; the one exception is step 6 (#342).
- Prefer current code and workflow files over older notes when they disagree.
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
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.

