Smart Wake
ReviewAudited by ClawScan on May 10, 2026.
Overview
This instruction-only skill is coherent for timeout recovery, but it teaches the agent to keep itself running through cron wakes and spawned sessions without an explicit user approval boundary.
Install only if you want agents to continue long-running tasks through scheduled wakeups and spawned resume sessions. Before using it, make sure you can approve, inspect, and cancel cron wakes, and avoid using it for sensitive or irreversible actions unless the resume scope is clearly controlled.
Findings (3)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
The agent may continue work later or off-hours without the user actively supervising that resumed session.
The skill instructs the agent to schedule future wakeups and spawn resumed sessions, which creates autonomous activity beyond the current interaction.
Cron wake to re-invoke agent at scheduled time... At wake time, call `sessions_spawn` (or `session_spawn`) with `wakeMode` enabling auto-resume via cron.
Require explicit user approval before creating cron wakes or spawned resume sessions, set a clear maximum runtime/wake count, and provide an easy way to view and cancel pending wakes.
A long-running task with sensitive or high-impact actions could be resumed automatically when the user did not explicitly approve continued execution.
The instruction encourages automatic activation of a scheduling and session-spawning workflow based on an estimate, without requiring user confirmation or limiting what kinds of tasks may be resumed.
If `estimated_duration > 70% timeout`: enable `smart-wake` immediately.
Add a consent gate and scope restrictions, especially before resuming tasks that write files, call external services, change accounts, publish content, or perform other irreversible actions.
If a checkpoint contains private information or is modified incorrectly, a future resumed session may expose that context or follow the wrong next step.
Persistent checkpoints are central to the skill, but later sessions rely on stored state and next actions, so checkpoint contents need to be trusted and should avoid unnecessary sensitive data.
Memory/slot for state persistence (state checkpoint)... Resume always reads latest checkpoint; never re-run blindly.
Store only the minimum necessary checkpoint data, avoid secrets, validate checkpoint origin and integrity before resuming, and show the checkpoint to the user when resuming important tasks.