Nonstop Mode is an autonomy-focused skill that enables session-scoped continuous work when the user is away. It enforces a mandatory pre-flight sequence (mental simulation, dangerous-ops manifest, permissions check, and explicit activation) to ensure the agent has no unresolved ambiguities before proceeding. During operation it applies a three-level blocker decision framework (solve, work around, mark blocked), spawns subagents for long-running tasks, and records dangerous operations so only pre-approved categories are executed. It also provides explicit deactivation steps and auto-deactivation safeguards to avoid runaway loops.
Use Nonstop Mode when the user requests the agent to "keep working" or be autonomous for extended periods, for example: unattended builds, long-running research, or multi-step automation where intermittent user input would be impractical. Do NOT use for actions that require sensitive approvals unless the user explicitly pre-approves dangerous categories.
Best paired with orchestrators, monitoring subagents, and agents that can safely perform infra or deploy tasks when explicit approvals are granted.
This skill has not been reviewed by our automated audit pipeline yet.
