Restart Task Recovery

Preserve and resume in-progress multi-agent work across OpenClaw config patch/apply restarts. Use when a restart is required during active tasks, when users...

This skill appears coherent and is designed to be a safe helper for resuming interrupted agent sessions, but review these before installing: - Confirm 'sessions_list', 'sessions_history', and 'sessions_send' are the legitimate platform tooling you expect — the scripts generate payloads that rely on those tools. - Check that checkpoint files (memory/restart-checkpoints) do not contain sensitive secrets or PII you don't want persisted; control access and retention of that directory. - The pre-resume verifier flags messages with words like 'retry', 'write', 'create', 'update', 'delete', 'payment', 'transfer' as high risk — verify that the manual confirmation workflow is enforced in your environment before auto-sending any held actions. - Test in a staging environment to ensure resume messages are idempotent and that automated 'send' actions don't trigger unwanted external writes. If you want greater assurance, run a code review of the scripts in your environment and validate that sessions_send calls behave safely under your policies.

Type: OpenClaw Skill Name: restart-task-recovery Version: 1.0.0 The skill bundle provides a legitimate framework for checkpointing and resuming multi-agent tasks during system restarts. It includes safety features like 'pre_resume_verify.py' to flag potentially non-idempotent actions (e.g., 'delete', 'transfer') for manual confirmation, and the scripts (e.g., 'build_checkpoint.py', 'recover_from_latest_checkpoint.py') are focused on structured data processing and local state management without evidence of malicious intent or data exfiltration.