What problem does it solve? Headless, scheduled server monitoring produces findings that nobody sees unless someone checks the machine. This Skill closes that gap by emailing the operator only when the delta of findings crosses a configured threshold, so a quiet machine stays quiet and real changes get attention. ## Core Features & Use Cases - Threshold-gated dispatch: Sends mail only when a new finding meets the minimum severity, a regression occurs, or a new outbound destination appears on a workstation, based on settings in config/watchman.conf. - Credential isolation: Reads SMTP credentials exclusively from .env through lib/smtp.sh, so secrets are never touched directly by the skill logic. - Graceful degradation: If mail is unconfigured or msmtp is missing, it logs and skips without crashing the monitoring loop. - Use Case: A cron-driven watchman run on a Linux server detects one regressed finding and two new high-severity findings; the skill builds a report from report-status output, appends the headless run-cost ledger, and emails the operator a subject like "watchman: 1 REGRESSED, 2 new high on host". ## Quick Start Ask the assistant to run the watchman loop and email me a report only if new high-severity findings or regressions appear on this server.
