Task Persistence
ReviewAudited by ClawScan on May 10, 2026.
Overview
This skill appears to persist task and session recovery data locally; the main things to notice are local snapshot storage and optional long-running monitoring.
Before installing, be aware that this skill is meant to leave recovery data on disk. Avoid snapshotting secrets, review or clean the workspace snapshot folders when needed, and only use the full monitoring mode if you want a process to keep running until stopped.
Findings (4)
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.
Potentially sensitive session context may remain on disk and be reused during later recovery.
Session snapshots can include conversation history, system state, active tasks, and pending operations, which are persisted to local JSON files for recovery.
"conversation_history": session_data.get("history", []) if include_context else [], "system_state": session_data.get("system_state", {}), "pending_operations": session_data.get("pending_ops", [])Use snapshots only for sessions you are comfortable persisting, review the snapshot directory, and clean old snapshots if they may contain secrets.
If intentionally started, the monitor may keep running and continue writing local gateway/task status until stopped.
The optional full mode starts a gateway monitor and then waits until interrupted, creating a long-running foreground process.
elif args.mode == "full": ... monitor.start_monitoring() ... signal.pause()
Use the documented check-restart/status commands for one-time checks, and only start full mode when you want a persistent monitor.
The agent may create or update local task records more often than a user expects, though this matches the task-persistence purpose.
The skill gives the agent broad situations where it may run local Python task-management commands and mutate task-state files.
Trigger on "resume", "任务恢复", "重启后", "未完成任务", or before/after any multi-step operation.
Check the task list/status periodically and ask the agent before registering sensitive or high-impact long-running tasks.
Users have less external context for verifying the publisher or comparing the package to an upstream repository.
The package does not provide a public source or homepage for independent provenance review, although the supplied code shows no hidden network or credential behavior.
Source: unknown; Homepage: none
Install only if you trust the registry publisher, and review the bundled scripts before relying on the skill for important task recovery.