Task Persistence

Task continuity, session snapshots, and gateway restart recovery. Use when starting long-running tasks (register them), after gateway restart (check for inte...

This skill's code matches its stated purpose (task persistence, snapshots, restart recovery) and does not contact external endpoints, but there are a few practical inconsistencies you should consider before installing: - Default workspace mismatch: SKILL.md shows /workspace as the default, but the Python code falls back to paths like ~/.openclaw/workspace or /home/admin/.openclaw/workspace if OPENCLAW_WORKSPACE isn't set. Set OPENCLAW_WORKSPACE explicitly to a safe directory to ensure files are written where you expect. - Docs vs code: dist/SKILL.md mentions optional dependencies (psutil, uv, memory-core) that the included scripts do not actually import; this looks like stale or copy-pasted documentation. Expect the packaged scripts to run with the standard library and python3. - File writes: the skill will create and modify files under the workspace (tasks/, memory/, persistence/, monitor/). If you have sensitive data in your workspace or care about disk usage, inspect the skill directory and run it in a sandboxed workspace first. - Background monitoring: running the 'full' mode starts a background monitor thread that will persist in the process. That is normal for this functionality but be mindful if you allow autonomous invocation. Recommended actions before enabling or granting autonomous use: 1. Inspect the packaged scripts (they are included) to confirm behavior (you've already been given them). 2. Explicitly set OPENCLAW_WORKSPACE to a dedicated sandbox directory. 3. Run check-restart/status commands manually to verify outputs. 4. If you want stronger assurance, run the skill in an isolated environment for a short period and confirm no unexpected files are written outside your chosen workspace. Overall: not obviously malicious, but internal inconsistencies and potentially surprising default paths justify caution.

Type: OpenClaw Skill Name: task-persistence Version: 1.1.0 The OpenClaw AgentSkills skill bundle 'task-persistence' is classified as benign. All Python scripts (`gateway_monitor.py`, `session_snapshot.py`, `task_persistence.py`, `task_queue.py`, `main.py`) consistently operate within a designated workspace, storing all state and task data locally. There is no evidence of data exfiltration, unauthorized network communication, or attempts to access sensitive system files outside the workspace. The `SKILL.md` instructions guide the AI agent to execute these scripts with well-defined arguments, showing no signs of prompt injection aimed at malicious actions. Minor inconsistencies, such as undeclared dependencies or differing execution commands (`python3` vs `uv run`) between `SKILL.md` versions, are present but do not indicate malicious intent.