sre-operationslisted

# Purpose Operate production responsibly: four SRE workflows with the same two invariants — evidence before action, and a human gate before anything destructive (rollbacks, restarts, scaling, config changes in shared envs). # The four workflows - **SLO definition** — user journeys → 3–5 SLIs (availability, latency p95/p99, throughput, correctness) → targets with error budgets (never 100%) → instrumented metrics → alerts, each linked to a runbook (an alert without a runbook is incomplete). - **Incident response** — classify severity; check the last 24h of deploys and the error logs in parallel; **human gate** before mitigation (fix or rollback); document the timeline; blameless post-mortem (5 Whys) within 48h. - **Capacity planning** — measure current utilization, analyze growth trends, forecast 3–12 months with confidence levels, recommend horizontal vs vertical with cost. - **Toil reduction** — inventory operational tasks by frequency × duration × automatability; automate the top items as scripts with safety checks and rollback (the `automation` skill carries the script standards). Platform specifics (runtime CLI, env aliases) come from project.json → `cloud`; never hard-code providers. Artifacts go to `{task-dir}/sre/{action}-{target}.md` when a task dir is in play. Details per workflow: `resources/workflows.md`. # How to use ## What it does Carries the SRE operating discipline: SLOs with error budgets and runbook-linked alerts, gated incident res