Observability AI Agent Directives

Observability is measured in the alerts that actually mean something and the log queries that do not bankrupt the team. These observability directives give an AI agent the specific job of producing those: generating Prometheus alert rules from a service's SLO, producing a Grafana dashboard as import-ready JSON, computing SLO burn rate and the multi-window thresholds that page proportionally, rewriting expensive log queries across Splunk, Datadog, and CloudWatch, and structuring an incident postmortem in the Google IRM format. The prompts are written to be pasted into Cursor, Windsurf, or Claude Code, so cursor rules prometheus, grafana, and slo rules land as prompts that produce rules YAML, JSON, and documents instead of prose. The unifying idea is that alerting and dashboards are code that should be generated and reviewed rather than clicked into existence. For anyone building a set of cursor rules observability or ai agent devops directives, this category closes the loop between an SLO an owner cares about and an on-call policy that actually fires. File the set together so alerting and dashboards become reviewed code instead of a hand-built artifact. They put the monitoring a team relies on into code review, so the alerts and dashboards annotate themselves with intent and survive whatever hands built them by clicking.