dsh-resume-on-restart

When the DSH service restarts, work-in-progress often grinds to a halt. Session context is lost, the agent faces a blank slate with no idea what happened or where to pick up. dsh-resume-on-restart was built for exactly this scenario—offering a lightweight bridge between a cold start and a productive resumption.

Its approach is deliberately simple. A small marker file records the previous process PID and boot time. On each DSH start, the plugin compares the new PID against the stored one to detect a restart. Once confirmed, it polls for the primary agent to become available (up to 20 seconds) and delivers an informational message summarizing the downtime and the most recently active session. Crucially, this is a notification, not a command—the agent decides for itself whether to resume prior work, adjust its plan, or start fresh. During normal operation, the plugin continuously tracks active sessions via cordis session-start and pre-step events, so the next restart notice carries meaningful context.

If you run multi-step agent workflows through DSH and do not want to manually reconstruct context every time the service recycles, this plugin is worth a look. It has zero dependencies on DSH internals, works inside the DSH Desktop packaged environment (app.asar), and offers sensible configuration knobs: disable waking, filter session ID prefixes, append custom notices, tune the grace window, and more.