What problem does it solve? Building event-driven background work for agents often couples side effects to the foreground chat request, making flows hard to deduplicate, trace, and extend. This Skill provides a consistent runtime model (source event → signal interpretation → action execution) so background agent work stays decoupled, idempotent, and observable. ## Core Features & Use Cases - Source, Signal, and Action Handlers: Register typed handlers with defineSourceHandler, defineSignalHandler, and defineActionHandler, then bundle them into installable policies. - Sync and Async Execution Entry Points: Choose between immediate execution, controlled backend execution, or queued out-of-band processing via Upstash Workflow. - Self-Iteration Receipts: Enqueue execAgent runs for memory and skill writes, then project user-visible receipts from the agent.execution.completed completion path. - Observability and Debugging: Use shared OTEL tracers, metrics, trace projection, and a development workflow snapshot bridge to inspect chains. - Use Case: When a user message signals feedback about memory, emit an agent.user.message source, route it through satisfaction and domain signal handlers, and enqueue a memory action whose receipt appears after the background run completes. ## Quick Start Use the agent-signal skill to add a new source handler and policy for a background agent workflow, then wire observability and tests.
