operations

Run the operations skill, which reads architecture docs, config files, and pipeline/IaC definitions to produce a 16-section runbook in docs/operations/. It outputs Markdown plus an offline HTML companion, flagging unknown values as ⚠ TODO rather than guessing.

The runbook covers at-a-glance summary, architecture and dependency maps, environments, access requirements, routine operations (deploy/rollback/restart/migrations), monitoring, secrets rotation, triage trees, failure-mode playbooks, backup/DR, escalation, and a command appendix.

Only in a limited, consent-gated way. Steps 1-3 use architecture docs, config, and IaC with no source reads; Step 4 may read up to 6 source files to confirm failure modes, each requiring explicit consent with justification.

Yes, the HTML companion is fully self-contained with inline CSS and no CDN links. Mermaid diagrams render only if a mermaid.min.js is vendored locally at docs/operations/vendor/; otherwise they degrade to readable source text.

The operations skill produces the ongoing operate-in-prod runbook used during incidents. The separate /go-live skill generates the one-time support-transition acceptance checklist that ingests readiness, security, and code-review ledgers.

TODO markers indicate values that require human or live-environment verification, such as on-call contacts, secret expiry dates, or SLA targets. The skill never fabricates these, since a wrong rollback step is more dangerous than a missing one.