proactive-agent

A proactive agent anticipates user needs through reverse prompting and scheduled check-ins. This Skill implements that behavior by adding self-improving routines and continuity safeguards, transforming a reactive model into a partner that initiates helpful actions autonomously.

Memory and continuity are preserved using a write-ahead logging (WAL) protocol that captures session state. The agent logs a Working Buffer during the danger zone before context limits are hit, then recovers critical details via Compaction Recovery on truncation.

Self-healing for long-running agents is best implemented via a heartbeat system. This routine continuously scans the agent's state, executes maintenance on its memory, and automatically recovers from failures or context truncation without manual intervention.

Security hardening is added by enforcing verify-before-reporting controls and injection defenses against unsafe external actions. The Skill includes a security audit script to validate the setup and ensure the agent operates safely within its autonomous scheduling boundaries.

Yes, bash is a required dependency for this framework. The Skill integrates scripts and assets into your agent workspace, using bash to execute the included security audit script and manage the underlying heartbeat and self-healing routines.

Session state is lost during compaction without a persistent logging mechanism. This Skill solves the problem by enforcing a write-ahead logging protocol and a Working Buffer exchange, ensuring the agent captures details before truncation and recovers them afterward.