Repository navigation
Epic 17 · Omarchy installs ncly with yay #457
Description
The problem (CMO)
On Linux, a user installs ncly by hand from the release archive. Omarchy is the reference Linux of nicely, but no package exists for it. On 2026-11-03, om1 ran a development build of 2026-10-28.
GoReleaser already builds the ncly-bin AUR package, and just release-check checks it. The upload stops on skip_upload: true (.goreleaser.yaml line 71). The secret AUR_SSH_KEY exists since 2026-10-06, as the private half of the key nicely-release-aur. No AUR account holds its public half, and the public half is not on om1.
Unknown: whether the AUR accepts new accounts now. The AUR idea before #442 said that registration was closed. A bot challenge blocked the check on 2026-11-03.
REF: COR-105 (docs/backlog/COR-105-aur-package.md)
The outcome (FMO)
Each nicely release publishes ncly-bin to the AUR. On Omarchy, yay -S ncly-bin installs the released ncly, and bash completes ncly <Tab>.
Stories
- No AUR account holds the release key. (chore(release): Pascal opens the AUR account for ncly-bin #448) · text
- Today: no AUR account holds a release key.
- After: Pascal's AUR account holds the public half of a new release key.
AUR_SSH_KEYholds its private half.
- Each release skips the AUR upload. (feat(release): each release publishes ncly-bin to the AUR #449) · code + tests
- Today:
skip_upload: truestops the upload. - After: the release publishes
ncly-bin. Therelease-nicelyskill and COR-105 say so.
- Today:
Done when
After the next release, yay -S ncly-bin on om1 installs the version of that release, and ncly <Tab> completes in bash (#449). The AUR page of ncly-bin shows that version.
Every Story has a documented disposition. The outcome has delivery evidence.
👨🏻🍳 For the agent
👨🏻🍳 Details
Out of this Epic
- A check of a stale
nclybuild inncly doctor: backlog issue feat(doctor): tell the user that the installed ncly is older than the latest release #454
Plan
#442 plans the AUR package as COR-105, third in the queue, with two cards: COR-105-T1 (Pascal) is #448, and COR-105-T2 (agent) is #449. Pascal can now do #448 after the merge of #442 through #459. Then an agent sets COR-105-T1 to done, and COR-105 becomes Active. just next serves #449 after COR-102 and XT-107 ship (#495 put XT-107 second). If COR-105-T1 is still todo at that time, just next names it as the card that waits on Pascal.
Blocked by
- feat(release): each release publishes ncly-bin to the AUR #449 is blocked by chore(release): Pascal opens the AUR account for ncly-bin #448, because an upload without an account fails the release.
- The dependency of chore(release): Pascal opens the AUR account for ncly-bin #448 and feat(release): each release publishes ncly-bin to the AUR #449 on docs(backlog): serve COR-104-T10 before COR-102, and make the tracking files true #442 is satisfied by 📚 docs: backlog: serve COR-104-T10 before Canadian French #459. COR-105-T1 and the COR-105 file now exist.
Approaches considered
- A PKGBUILD written by hand: refused. GoReleaser already builds and checks the package.
- An update notice like the one of
gh: refused. The principles allow network access only on request. - Linuxbrew only: refused (Pascal, 2026-11-03, 4a).
- Reuse of the old key: refused. Its public half is lost on om1, and nobody can compare a secret with a fingerprint.
by Opus-5.5 via Claude Code
Activity