agentleFS
silo-codesilo-code/silo/.agents/skills/silo-release-publish/SKILL.md

silo-release-publish

Skill62 starsChanged 59 days agoMITView on GitHub

Publish a drafted Silo app release on GitHub — verify the draft's assets, expand its notes to cover every unpublished version back to the last public one, flip it to public, and confirm the updater picks it up. Use when asked to publish/ship a release, "did we ever publish X", or when a release tweet or announcement needs to match what users can actually install.

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 3 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

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

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

    Do not publish unless all of these are present and uploaded:

    line 40
  2. Very specific:

    Read the result before applying it — check the callout names the right versions and the awk range didn't clip or duplicate a section.

    line 79
  3. Distinctive:

    All three must agree on the version you just published.

    line 101
  4. Distinctive:

    Always produce a tweet draft as the final step — the user shouldn't have to ask.

    line 107
  5. Distinctive:

    No hashtags, no links, no @-mentions.

    line 133

≈ “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

Publishing a Silo release

Silo app releases land as drafts on purpose — a soak-test checkpoint before they go public (see apps/docs/guide/release-channels.md). Publishing is a separate, manual step, and it is the moment the release becomes real: GitHub's "Latest" slot, the website download button, the docs download links, and every existing install's auto-updater all follow /releases/latest immediately.

Because drafts accumulate silently, the version you're asked to publish is often not one version ahead of what users are running. Always establish the gap before writing notes.

Package releases (sdk-*, git-api-*) are pre-releases and are published by CI. This skill is only for silo-v* app releases.

1. Establish the gap

gh release list --limit 10
gh api repos/silo-code/silo/releases --jq '.[] | select(.draft==false) | select(.tag_name|startswith("silo-")) | .tag_name' | head -3

The first published silo-v* tag is what users are actually on. Every draft between it and the one you're publishing is a version whose changes ship silently unless you fold them into the notes.

2. Verify the draft is complete

gh release view silo-vX.Y.Z --json name,isDraft,isPrerelease,tagName,assets \
  --jq '"draft:\(.isDraft) pre:\(.isPrerelease)", (.assets[] | "\(.name)  \(.size)  \(.state)")'

Do not publish unless all of these are present and uploaded:

  • latest.json — the updater manifest
  • macOS: Silo_X.Y.Z_aarch64.dmg, Silo_X.Y.Z_x64.dmg, Silo_aarch64.app.tar.gz(+.sig), Silo_x64.app.tar.gz(+.sig)
  • Windows: Silo_X.Y.Z_x64-setup.exe(+.sig), Silo_X.Y.Z_x64_en-US.msi(+.sig)
  • Linux: Silo_X.Y.Z_amd64.AppImage(+.sig), Silo_X.Y.Z_amd64.deb(+.sig), Silo-X.Y.Z-1.x86_64.rpm(+.sig)

A missing .sig means the updater will reject the build for that platform — stop and investigate rather than publishing a partial release.

Confirm latest.json names the version you expect:

gh api -H "Accept: application/octet-stream" \
  "$(gh release view silo-vX.Y.Z --json assets --jq '.assets[]|select(.name=="latest.json")|.apiUrl')" \
  | python3 -c "import json,sys; print(json.load(sys.stdin)['version'])"

3. Expand the notes to cover the gap

release-please writes each draft's body from that version's commits alone. When intermediate drafts were never published, that body understates the release — users jumping several versions would see only the last one's changes.

Build the body from the changelog, which already has every section in the right format. Prepend a callout naming the gap, keep the target version's own body, then append each skipped version's section verbatim:

{
  echo "> **This is the first public release since vA.B.C.** It also includes everything from vD.E.F and vG.H.I, whose builds were never published. All of those changes are listed below."
  echo
  gh release view silo-vX.Y.Z --json body --jq .body
  echo
  # every changelog section from the newest skipped version down to (not incl.) the last published one
  awk '/^## \[D\.E\.F\]/,/^## \[A\.B\.C\]/' apps/desktop/CHANGELOG.md | sed '$d'
} > /tmp/notes.md

Read the result before applying it — check the callout names the right versions and the awk range didn't clip or duplicate a section. Skip the callout entirely when there is no gap (the previous version is already public).

4. Publish

gh release edit silo-vX.Y.Z --notes-file /tmp/notes.md --draft=false --latest -R silo-code/silo

--latest is explicit on purpose: it guarantees the app release takes the "Latest" slot rather than a package pre-release.

5. Confirm it went live

gh release view silo-vX.Y.Z --json isDraft,isPrerelease,publishedAt
gh api repos/silo-code/silo/releases/latest --jq .tag_name
curl -sL https://github.com/silo-code/silo/releases/latest/download/latest.json \
  | python3 -c "import json,sys; print(json.load(sys.stdin)['version'])"

All three must agree on the version you just published. The last one is the actual updater endpoint — if it still reports the old version, existing installs will not see the update.

6. Draft the release tweet

Always produce a tweet draft as the final step — the user shouldn't have to ask. The "what's new" story is everything since the last published version (step 1), not just the target tag.

Format (match it exactly — this is the house style for @silo_code):

Silo vX.Y.Z is out 🚀

- <headline feature — concrete, user-facing, one line>
- <headline feature>
- <headline feature>

Fixes: <short comma-separated list of the notable fixes>

Rules:

  • Three headline bullets, max four. Pull them from the Features sections of every version in the gap, not just the target. Pick what a user would notice — collapse a multi-phase RFC (e.g. Agent Profiles phases 1–2) into one bullet.
  • Phrase bullets as an outcome, not a changelog line: "Git branches modal: keyboard nav + folder switcher", not "git-explorer: branches modal keyboard nav". Drop package/scope prefixes and issue numbers.
  • One Fixes: line, comma-separated, plain language. Skip internal-only fixes (test flakes, release-please manifest resyncs, CI).
  • No hashtags, no links, no @-mentions. Keep it under ~280 chars if you can, but readability wins over the limit.
  • Present it in the reply as a fenced block the user can copy. Don't post it.

What publishing does not fix

  • The in-app update prompt renders the notes field baked into the signed latest.json asset, which still holds only the target version's own notes. Expanding the GitHub body does not change it. Say so when reporting the publish; only re-upload that asset if the user asks.
  • Older drafts stay drafts. They're harmless, but the version history will have gaps. Offer to publish them as backfill; don't do it unprompted.

Reporting back

Lead with the release URL and confirmation that the updater endpoint serves the new version. Then give the tweet draft from step 6, then name what's still outstanding — the in-app prompt's notes, any remaining drafts. If a separate release announcement is already in flight, check it still matches: the "what's new" story is everything since the last published version, not since the last tag.

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.