Schedule the agent, not the script

A traditional cron job runs /usr/local/bin/reconcile.sh at 2am. The script is fixed. If reconciliation needs different logic on month-end versus a regular Tuesday, you write that branching into the script and pray the edge cases hold.

A Hoody cron entry can curl a hoody-agent URL with a prompt: "Reconcile yesterday's transactions. Today is the last day of the month, so include the monthly close. Notify me of anomalies."

The agent reads the date, reads the data, decides what to do. The schedule is fixed; the work isn't. You write the prompt once. Edge cases stop being branches you maintain: they become context the agent reasons about.

The cron entry doesn't run the job: it asks an agent to figure out the job.