runlisted

# /rein:run Thin orchestration that executes **one** task and records evidence. Selection, ordering and closure live in the `rein` CLI, not in your judgement. **One task per invocation. No `--all`.** Usage: `/rein:run [change-name]` Resolve the CLI once and reuse it: ```bash R=$(command -v rein || ls -d ~/.claude/plugins/cache/*/rein/*/bin/rein 2>/dev/null | sort -V | tail -1) ``` ## Steps 1. **Record this invocation** — never blocks, never fails the run. Shell state does NOT persist between tool calls, so `R` is resolved and used in the SAME block or it is empty and nothing is recorded: ```bash R=$(command -v rein || ls -d ~/.claude/plugins/cache/*/rein/*/bin/rein 2>/dev/null | sort -V | tail -1); "$R" event run ``` 2. **Ask the gate, do not choose yourself**: `"$R" next .` — it returns `{ready, taskId, humanReview, verification, reason}` and exits non-zero when nothing is claimable. If `ready` is false, report `reason` and **stop**. 3. If `humanReview` is true, **stop before claiming it** and tell the user this task needs their judgement. 4. Show the selected task and its acceptance criteria (`"$R" tasks .`). Load **only** what that task needs — the criteria, the relevant design notes, and the files it touches. Read symbols and regions, never whole files speculatively: every file you pull in is re-read on every later turn. 5. State a brief implementation plan, then implement **strictly within scope**. 6. Run the task's own `Verificatio