sue-cronjob

What problem does it solve?

Managing recurring scheduled tasks for SUE scale-up ML/HPC experiments is error-prone when juggling multiple experiment runs across different backends, as manual cron edits can accidentally disrupt unrelated system jobs or miss critical monitor and cleanup triggers.

Core Features & Use Cases

• Tagged Cron Safety: Only modifies cron entries explicitly tagged with SUE_CRONJOB id=<id> to avoid accidental changes to unrelated scheduler jobs.
• Multi-Platform Support: Handles local Unix cron, Codex CLI cron wrappers, Kimi CronCreate reminders, RunPod cleanup cron, and macOS LaunchAgents for cluster snapshot jobs.
• Guardrail Enforcement: Enforces default Asia/Shanghai timezone, secret redaction, disable-instead-of-delete for temporary stops, and mandatory dashboard observability via Turso status writes.
• Use Case: When running a long SUE fullrun experiment on RunPod, use this skill to set up a 10-minute recurring monitor cron that alerts you if GPU hours are missing, without risking other system cron entries.

Quick Start

Use the sue-cronjob skill to list all active SUE-tagged cron jobs for your current experiment workspace.