watch-cilisted

# watch-ci The standalone entry for CI watching: same watch-and-triage flow as the pipeline's watch-ci stage, reached directly instead of through a unit of work. The target comes from the conversation and the checkout, and the verdict goes to the user. ## Run Outside a pipeline this verb is its own run, so the console shows it and the Stop hook covers its pane. Skip this section when `RT_RUN_DB` is set and `rt runs snapshot` shows `run.status` = `running`: you were invoked from inside that run, you inherit it, `run.current_stage` is your stage, and you close nothing at the end. Otherwise, when a surface launched this pane (the `--spawned-by` case below), start fresh: another pane's live run is not yours to resume. Launched by hand, first the Resume offer: list `~/.mattstack/runs/<repo>/` (the `--repo` value in the flags block below) for runs whose `snapshot` shows `run.status` = `running` and `run.work_type` = `watch-ci` (read each with `RT_RUN_DB` pointed at its `state.db`; never raw sqlite). Any found: gate `clarify`, one sentence naming each candidate's `spawned_by`, `started_at`, and `current_stage`, then the structured-question tool with one **Resume** option per candidate (recommended for a run this session started earlier; a run another live pane owns is not yours) / **Start fresh**; **Hold**. Resume: `export RT_RUN_DB=<its state.db>`, then `rt runs stage-start --stage watch-ci` (a new attempt, which re-records this session) and `rt runs field set hold - --stage wa