evening-review

To generate an end-of-day recap for a multi-agent system, trigger a daily cron-driven workflow that aggregates completed tasks, heartbeats, and inbox items into a consolidated daily report. This process also evaluates agent performance and proposes overnight work.

Orchestrator self-evaluation in agent orchestration is a process that rates multi-agent performance across usefulness, proactivity, coordination, communication, and learning. It generates actionable proposals to fix bottlenecks or automate slow steps within the system.

To dispatch overnight tasks with user approval in a PTY-based agent fleet, the system proposes autonomous overnight work following nighttime-mode guardrails. It uses a Telegram approval flow to ensure dispatch behavior is strictly approval-only before executing tasks.

Yes, you can use Telegram approvals for task planning in autonomous agent workflows. The system integrates a user approval flow to review proposed overnight work, ensuring that task dispatch behavior strictly requires your explicit approval before execution.

Yes, the daily recap workflow persists memory updates across all agents. It reads identity, goals, and nighttime-mode constraints, aggregates the bus state including tasks and heartbeats, and saves memory updates during the end-of-day recap generation.

You should enforce nighttime-mode constraints for task dispatch planning when proposing autonomous overnight work. This ensures that all dispatched prep tasks for tomorrow follow strict guardrails and operate safely under an approval-only behavior model.