night-modelisted

<!-- @harness-posture: portable-designed-not-proven — designed for any harness with sequential subagent dispatch (SKILL.md:19); the end-to-end non-CC run is NOT yet exercised — that honest gap stays open until a live probe --> > **Authoritative for:** the OVERNIGHT DELTA on top of `subagent-driven-development` — the unattended autonomy/fork policy, quota/backoff resilience, Workflow context-economy, verification discipline, and the terminal condition for an unsupervised run. > **NOT authoritative for:** the executor + review dispatch loop itself — that is `superpowers:subagent-driven-development` (SSOT #64, ADOPT), which this skill **layers over, never re-describes**. Project goal — see [README.md#why-this-exists](../../../README.md#why-this-exists). # /night-mode — thin overnight layer over subagent-driven-development Run a well-scoped implementation task to completion with **no human in the loop**. The dispatch loop — a coordinator dispatching a fresh implementer per increment, then SDD's review seats, with a rework loop to convergence — **is `superpowers:subagent-driven-development` (SDD)**. Use it. This skill adds **only** what SDD does not cover for an _unsupervised overnight_ run. If you catch yourself re-describing the executor/reviewer roster or the review loop, stop — that is SDD's job, and re-describing it is the `#parallel-evolution-creep` this project exists to prevent. ## The loop is SDD (do not reinvent) Invoke `superpowers:subagent-driven-development` for