← ClaudeAtlas

scheduled-automation-runtimelisted

Local scheduler jobs need proof when launchd, systemd, cron, or Windows Task Scheduler differ from manual runs or lack runtime proof. Not for vendor CLI command construction, architecture inventory, cloud schedulers, or job business logic.
Xopoko/plug-n-skills · ★ 11 · AI & Automation · score 77
Install: claude install-skill Xopoko/plug-n-skills
# Scheduled Automation Runtime Use this skill to prove what a local operating-system scheduler actually ran. A successful interactive-shell run is inner-command evidence, not scheduler evidence. Resolve `$PLUGIN_ROOT` from the host when available; otherwise use the absolute path of this skill folder's `../..`. Load only the relevant reference: - launchd: `$PLUGIN_ROOT/references/launchd-runtime-proof.md` - systemd timers: `$PLUGIN_ROOT/references/systemd-timer-runtime-proof.md` - Windows Task Scheduler: `$PLUGIN_ROOT/references/task-scheduler-runtime-proof.md` - cron: `$PLUGIN_ROOT/references/cron-runtime-proof.md` - P3/P4 proof, canaries, receipts, or repair: `$PLUGIN_ROOT/references/scheduled-run-proof-contract.md` ## Evidence Ladder Keep these claims separate: - **P0 - defined:** the job definition exists and parses. - **P1 - registered:** the intended scheduler domain accepted the job, with enablement or armed state recorded separately. - **P2 - demand-run:** the scheduler manager can start it with the saved runtime context. - **P3 - triggered:** the configured native, non-demand trigger condition fired and produced a current correlated run. - **P4 - effective:** the intended functional effect succeeded. Only P3 proves the bound job's scheduling. A demand start proves at most P2. A separate canary can prove its own P3 and copied runtime properties, but can never promote the production job above its independently observed level. ## Workflow ### 1. Bind