playbook
Ops Automation Health Check
Automation inventory audited, credentials and integrations verified, safe patches applied, and external side effects gated before close
Monthly ops review, before patching scheduled jobs/integrations, or when an automation silently fails
How it runs
- scheduled-jobs → list managed jobs and compare against the live scheduler without writing changes
- tasknotes → inventory scheduled-run TaskNotes: identify enabled recurring tasks, `run_command`, `verify_path`, `max_turns`, model, lock settings, and last output
- Credential and integration check: verify required tokens/configs exist and identify expired, revoked, missing, wrong-account, or over-scoped credentials
- API health check: run non-mutating health probes for each integration and inspect recent logs for repeated failures, retries, or silent no-ops
- Side-effect gate review: classify every automation as read-only, local-write, vault-write, or external-write; confirm outbound actions are disabled, dry-run capable, or explicitly approved before patching
- Patch stale scripts/configs: apply the smallest safe changes, one automation family at a time, with backups where the tool supports them
- Verification and audit close: rerun health checks, verify expected outputs, capture decisions, unresolved risks, and next scheduled review
What it needs
- Current scheduled jobs manifest: `10-ME/00-PROFILE/scheduled-jobs.yaml
- Scheduled-run TaskNotes under `00-CORE/TaskNotes/Tasks/`
- Recent scheduled-run logs under `99-TMP/OUTPUT/scheduled-runs/`
- Relevant skill configs and `.env` files for integrations being checked
- Known external systems: Google Workspace, YouTube, Gmail, Google Drive
What you get
- Ops health audit note with checked systems, findings, fixes, and resid
- Updated scheduled-jobs manifest or TaskNotes only where changes were n
- Patch summary with files changed and verification evidence
- Escalation list for credentials, provider outages, or owner decisions