Live: Buzz #infra alert when updates stop (Cron Trigger) #9

Goal. When live.setkernel.com stops receiving updates for more than 5 minutes, post once to Buzz #infra, and again when updates resume.

Work

• A Cron Trigger (every minute) in wrangler.jsonc checks the Durable Object's lastIngest. Only one alert per outage, plus one recovery message.
• Posting needs the Worker's own Buzz identity: a Nostr key. Sign with a small audited library such as @noble/secp256k1, pinned. The key lives in Infisical (project setkernel-live, prod) and reaches the Worker as a secret through the deploy workflow; it is never in code or logs. Document the owner steps: generate the key without printing it, add it as a relay member, put it in Infisical.
• A mute switch (a wrangler.jsonc var) for planned maintenance.
• Buzz runs in the office, so a full office outage can't be reported while it lasts. Post on recovery with how long it was down.
• Tests: staleness detection, one alert per outage, recovery, mute.

Done when npm run validate passes and the PR documents the owner steps.

Decision: setkernel/hq#113 (Raghav's answers in its comments). Tracking: setkernel/hq#105.