Fix Everything Triage: label Omarchy issues

Fix Everything Triage sorts Omarchy's open issues into queues — look now, ready for an agent, upstream, ask the reporter, park — with TypeSafe's Jev model making the judgment calls. It is read-only: it never labels, comments on or closes anything. Whether it sorts well can only be checked against people's judgment of the same issues, so here are 80 open issues drawn at random across ages, lengths and labels. Label as many as you like, even one, then press send my labels: the page opens an issue on this project's repository with your answers filled in, or you can paste them as a reply on the Fix Everything discussion. Handing this to an agent? Point it at AGENTS.md, and say in the issue that an agent answered.

so your labels can be told apart from someone else's; kept only in this browser

Four questions per issue, answered from the issue as written. Each body is rendered the way GitHub shows it (images are links, so nothing loads from outside this file); show raw markdown switches to the text as written. Your answers save in this browser as you go; export downloads them. Keys: 1–5 kind · a/s actionable yes/no · j/k ahead of the backlog yes/no · d/f check privately yes/no · ? can't tell (for the question you last touched) · ←/→ move. Judge the issue as a maintainer would, not what a tool might say about it.

Once you have copied your labels, compare with a model's answers unlocks: one model's labels and notes for all 80, shown beside yours for the kind and actionable questions (it answered an earlier wording of the third question and had no fourth). The triage is graded against people's labels; the model's are there to compare with. You can keep labelling afterwards, and anything you answer after opening them is marked seen=model in your block so it can be counted separately.

Send these as an issue, or paste them as a reply on the Fix Everything discussion

It lists only the issues you answered. ? means you picked "can't tell"; seen=model means you answered after opening the model's answers.