proactive-agent

AI agent memory integrity is maintained using the WAL Protocol and a Working Buffer to capture context and enforce data consistency across heartbeat routines. This prevents context loss during long-running projects by systematically preserving session state.

The WAL Protocol is a memory management mechanism that logs context changes before applying them, ensuring data integrity during compaction recovery. It works with the Working Buffer to safely persist agent state across session interruptions.

AI agents are secured against prompt injection through automated security audits and security hardening features. These defenses scan for malicious inputs and enforce strict boundaries to protect long-running automation workflows from compromise.

Yes, you can automate onboarding for self-improving AI lifecycles using an onboarding-aware architecture with ONBOARDING/USER/SOUL collaboration. This maintains context and triggers growth loops to expand agent capabilities proactively.

To start heartbeats for a proactive AI agent, place the skill files in the openclaw/proactive-agent directory and initialize the heartbeat routines. This begins proactive operation, enabling self-healing workflows and continuous context capture.

Compaction recovery limitations depend on the Working Buffer capacity and heartbeat frequency during memory flush operations. If buffer limits are exceeded or heartbeats fail, context preservation across long-running sessions may be incomplete without manual intervention.