work-once

Run one skill for one issue, write the returned next skill onto that issue, and exit. Use when a schedule wakes an agent, when the user runs /work-once, or when one tick should resume from the issue tracker. Prefer an open issue that already has next_skill and is not waiting on a person; a person unblocks those with backlog, which this skill never runs. Otherwise run fix-loop on the oldest actionable issue, one issue only. Do not load the returned next skill in this run. Do not use to burn down a backlog in one sitting (fix-loop), to plan a replacement (yak-shave), to perform parallel change (expand-and-contract), or to file a defect (bug-report), except as the single skill this tick was told to run.