small cron watchdog that catches silent pipeline death. a pipeline can look healthy while data quietly stops flowing, so this checks the flow itself instead of process health.

it checks three things. newest file age in a data directory, newest file age in a nightly job output directory, and a health endpoint that should return ok json.

if anything is stale or unreachable it writes a small status json, prints the problem to stderr and exits 1. cron mails stderr on failure so problems get loud. success is silent so there is no mail spam.

edit the config block at the top with your paths and health url, then run it hourly from cron. stdlib only, no dependencies.