What problem does it solve? Recurring and scheduled work (digests, cleanups, periodic notifications, end-user-defined schedules) cannot rely on in-process timers like setInterval or node-cron because Cloud Run terminates idle instances. This Skill provides the correct patterns for implementing durable, platform-managed cron jobs on Manus webdev fullstack and mobile-app projects. ## Core Features & Use Cases - Heartbeat (HTTP cron): Platform POSTs directly to /api/scheduled/* handlers on your deployed site, with tRPC mutations for end-user-driven schedule creation and SDK functions (createHeartbeatJob, updateHeartbeatJob, deleteHeartbeatJob) for lifecycle management. - AGENT cron: Spawns a fresh isolated Manus agent per trigger for work needing agentic capabilities like web browsing, deep research, or image generation, with curl-based write-back to the site's scheduled endpoints. - Sandbox CLI management: Use manus-heartbeat commands to create, update, pause, resume, delete, list, and inspect logs of project-level crons owned by the project owner. - Use Case: A marketing site needs a daily 9am campaign email send. You add a schedule_cron_task_uid column, create a tRPC mutation calling createHeartbeatJob, mount an Express handler at /api/scheduled/sendMarketing that authenticates via sdk.authenticateRequest and looks up the campaign by taskUid, then deploy so end-users can schedule their own sends. ## Quick Start Read this Skill first, then ask the agent to set up a daily scheduled job on the Manus site by creating a tRPC mutation, an /api/scheduled/ handler, and deploying before scheduling.
