goal-release — Claude Code skill: release from PR to Homebrew
Drives a release end to end: opens the PR, waits for green CI, merges it and updates the Homebrew formula. Handy when a version has to ship in one go.
npx skills add dyoshikawa/rulesync --skill goal-release -a claude-codeWhat goal-release does
The skill chains other skills together. First draft-release creates a release/v<version> branch with the version-bump commits, opens a pull request against main and creates a draft GitHub release with notes. Then goal-release waits on the GitHub Actions checks, fixes failures on the release branch and pushes, and once CI is green calls merge-pr, which re-verifies the PR state and merges with gh pr merge --admin --merge.
You can pass a version or leave it out. A supplied new_version has to match ^v?\d+\.\d+\.\d+$, otherwise the skill stops and reports instead of continuing. With no argument the version comes from release-dry-run. CI fixes are capped at 3 attempts, and they have to be real fixes — skipping or deleting tests, weakening lint or type-check config, or editing workflows is off the table.
The last step is the Homebrew formula. The skill waits for the Publish Assets workflow, then for the release to go public (while it is still a draft, asset URLs return 404 and brew install breaks for everyone), downloads SHA256SUMS and regenerates Formula/rulesync.rb with scripts/generate-homebrew-formula.ts. That change lands as its own PR. You need gh, pnpm and admin rights on the repository.
What it helps with
- ship a release from version bump to merge without manual steps
- wait for green CI and fix failing checks on the release branch
- regenerate the Homebrew formula from fresh checksums
- avoid merging the formula before the release goes public
- make sure the merged PR is the release one, not some other
Example prompts for the agent
- cut a release v1.2.3
- ship the next version end to end, pick the version yourself
- run goal-release and update the Homebrew formula
How to install goal-release
1Skills CLI (Vercel)
npx skills add dyoshikawa/rulesync --skill goal-release -a claude-code2Manually: put the files in place
~/.claude/skills/goal-release/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/goal-release/
1Skills CLI (Vercel)
npx skills add dyoshikawa/rulesync --skill goal-release -a codex2Manually: put the files in place
.agents/skills/goal-release/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add dyoshikawa/rulesync --skill goal-release -a cursor2Manually: put the files in place
.cursor/skills/goal-release/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install dyoshikawa/rulesync goal-release2Manually: put the files in place
.github/skills/goal-release/SKILL.md1Skills CLI (Vercel)
npx skills add dyoshikawa/rulesync --skill goal-release -a gemini-cli2Platform CLI
gemini skills install https://github.com/dyoshikawa/rulesync --path .rulesync/skills/goal-releaseManually: put the files in place
.devin/skills/goal-release/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add dyoshikawa/rulesync --skill goal-release -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about goal-release
How do I install goal-release in Claude?
Run npx skills add dyoshikawa/rulesync --skill goal-release -a claude-code. The skill lands in ~/.claude/skills/goal-release/SKILL.md.
What does goal-release do?
It carries a release through to the end: opens the release PR and draft GitHub release via draft-release, waits for CI to pass, merges the PR with the merge-pr skill, then rebuilds the Homebrew formula from the published assets.
What does the skill need to work?
gh and pnpm installed, plus admin rights on the repository, since the merge runs through gh pr merge --admin --merge. If you pass a version yourself, it must match ^v?\d+\.\d+\.\d+$.
Source text
SKILL.md · MIT
---
name: goal-release
description: >-
Cut a release end to end: use the `draft-release` skill to open the release PR and draft
GitHub release, wait for CI to turn green, run the `merge-pr` skill to merge
the release PR, then update the Homebrew formula from the published release
assets. Use when the user wants to draft and merge a release in one go, or
triggers on "the `goal-release` skill".
targets:
- "*"
---
# Goal Release
new_version = the user's request
This skill drives a release all the way to merge. It uses the `draft-release` skill to
open the release pull request (and create the draft GitHub release), waits for
the PR's CI checks to pass, runs the `merge-pr` skill to merge it, and finally updates
the Homebrew tap formula once the release assets are built.
## 1. Draft the Release
Use the `draft-release` skill with `new_version`.
- If `new_version` is provided (e.g. `v1.2.3` or `1.2.3`), it must match the
semver shape `^v?\d+\.\d+\.\d+$` — if it does not, stop and report instead
of passing it on. Pass a valid value through unchanged; the `draft-release` skill
normalizes the `v` prefix itself.
- If `new_version` is empty, pass no argument; the `draft-release` skill determines the
next version automatically via the `release-dry-run` skill.
When the `draft-release` skill finishes, it has:
- created a `release/v<version>` branch with the version-bump commits,
- opened a pull request against `main`, and
- created a draft GitHub release `v<version>` with the release notes.
If the `draft-release` skill failed partway (e.g. the PR exists but the draft release
was not created, or it stopped before opening the PR), stop here and report
the partial state to the user instead of continuing.
## 2. Resolve the Release PR
Use the PR number or URL that the `draft-release` skill's `gh pr create` printed in
Step 1 when available. Otherwise, identify the pull request from the current
`release/v<version>` branch:
```bash
gh pr view --json number,title,state,headRefName
```
Confirm that `headRefName` matches the `release/v<version>` branch created in
Step 1. This skill must only ever merge that release PR — if the resolved PR
is a different one, stop and report to the user instead of merging.
## 3. Wait for CI
Wait for the release PR's GitHub Actions checks to finish:
```bash
gh pr checks <pr_number> --watch
```
If the watch reports that no checks are registered yet, wait a moment and
retry — checks can take a few seconds to appear right after the PR is opened.
- If every check passes, proceed to Step 4.
- If a check fails, investigate and fix the failure on the release branch
(run `pnpm cicheck` locally, commit, and push), then wait for the re-run.
Never proceed to the merge while any check is `fail` or `pending`.
- Fixes must legitimately resolve the failure. Never make a check green by
skipping or deleting tests, weakening lint or type-check configuration, or
editing GitHub Actions workflows.
- Set a safety cap of **3** fix attempts. If CI is still red after the cap,
stop and report the failing checks to the user instead of merging.
If any fix commit beyond the `draft-release` skill's own version-bump commits was
pushed, review the extra commits yourself instead of stopping: merge when each
one is a legitimate CI fix (it does not skip or delete tests, weaken lint or
type-check configuration, touch `.github/**`, or add a runtime dependency) and
CI is green, and list the extra commits in the final report. If any extra
commit fails that check, stop and report it instead of merging.
## 4. Merge the Release PR
Use the `merge-pr` skill with the release PR number. It re-verifies the PR
state and CI status, merges with `gh pr merge --admin --merge`, posts a
thank-you comment, and cleans up the local branch.
Note: the `goal-pr` skill tells agents not to auto-merge PRs that touch `package.json`
or release configuration. That restriction does not apply here — merging the
version-bump release PR is this skill's explicit purpose, and the user opted
into it by invoking the `goal-release` skill.
## 5. Update the Homebrew Formula
After the release PR merges, the `Publish Assets` workflow builds the platform
binaries, tags the release, and uploads the assets (including `SHA256SUMS`) to
the draft GitHub release. The Homebrew formula embeds those checksums, so it
can only be regenerated once that workflow finishes — this step replaces the
old `homebrew` job that `.github/workflows/publish.yml` used to run.
1. Wait for the `Publish Assets` run triggered by the merge to complete:
```bash
gh run list --workflow "Publish Assets" --branch "release/v<version>" \
--limit 1 --json databaseId,status,conclusion,headBranch
gh run watch <run_id>
```
The workflow triggers on every closed PR to `main` (non-release runs are
job-level skipped but still listed), so filter by the release branch as
shown. The run can take a few seconds to appear right after the merge — if
the list is empty, wait and retry. Confirm the watched run belongs to this
release (its `headBranch` is the `release/v<version>` branch of the PR
merged in Step 4) and concluded with `success`. If it failed, stop and
report — the formula must not be regenerated from stale assets.
2. Wait until the GitHub release is published (no longer a draft). The
`Publish` workflow (`.github/workflows/publish.yml`) runs after
`Publish Assets`, publishes the npm package, and then flips the release
public. Draft-release asset URLs return 404 for unauthenticated users, and
the Homebrew tap serves the formula straight from `main`, so merging the
formula before the release is public would break `brew install` for
everyone:
```bash
gh release view v<version> --json isDraft --jq .isDraft
```
Poll until this prints `false`. If the `Publish` workflow failed (e.g. npm
publish error) and the release stays a draft, stop and report instead of
updating the formula.
3. Regenerate the formula from the released checksums:
```bash
git checkout main && git pull
gh release download v<version> --pattern SHA256SUMS --dir ./tmp --clobber
pnpm exec tsx scripts/generate-homebrew-formula.ts <version> ./tmp/SHA256SUMS Formula/rulesync.rb
rm -f ./tmp/SHA256SUMS
```
4. If `git diff --quiet -- Formula/rulesync.rb` reports no change, the formula
is already up to date — skip ahead to the final report.
5. Otherwise commit the regenerated formula on a branch and merge it right
away (`main` is branch-protected, so the change must land via a PR; merging
it immediately with admin rights is this skill's explicit purpose, same
as the release PR itself):
```bash
git switch -c homebrew-formula/v<version>
git add Formula/rulesync.rb
git commit -m "chore: update Homebrew formula to v<version>"
git push -u origin homebrew-formula/v<version>
gh pr create --base main --title "chore: update Homebrew formula to v<version>" \
--body "Automated formula update for the v<version> release."
gh pr merge --admin --merge --delete-branch
```
If a previous attempt already pushed the `homebrew-formula/v<version>`
branch or opened its PR, reuse them instead of failing: force-push the
branch with `git push --force-with-lease` and skip `gh pr create` when
`gh pr list --head homebrew-formula/v<version>` shows an open PR.
Only `Formula/rulesync.rb` may be committed here. If anything else shows up
in `git status`, stop and report instead of committing it.
## 6. Final Report
Report to the user:
- The merged release PR number and title.
- The new version and a link to the GitHub release (published automatically by
the `Publish` workflow, which Step 5 waits for).
- Whether the Homebrew formula was updated (and the formula PR number) or was
already up to date.
- Any CI fixes that were needed along the way.
© dyoshikawa · MIT · dyoshikawa/rulesync