Skill pack — Specialist

Observability & Incident Response

OpenTelemetry, SLOs, alerts that only fire when it matters, and calm, well-run incidents.

What you get

incident-commander131 lines
Runs production incidents calmly from declaration to post-mortem: severity, roles, single channel, timeline, mitigation before root cause, status page and stakeholder updates, handover, closure and blameless review with owned actions. Use when an outage or degradation is happening, when deciding whether to declare, drafting incident updates, or writing and reviewing a post-mortem.
observability-engineer144 lines
Instruments services with OpenTelemetry traces, metrics and structured logs that answer real production questions at a sane cost. Use when adding telemetry to a service, setting up the OTel SDK or Collector, choosing sampling, fixing missing or broken traces, cardinality or cost blow-ups, uncorrelated logs, PII in telemetry, or building service dashboards.
slo-alerting-engineer125 lines
Defines SLIs and SLOs from user journeys, sets error budgets and policy, and builds multiwindow multi-burn-rate alerts that page only when users are hurt. Use when creating or reviewing SLOs, writing Prometheus or vendor alert rules, reducing alert fatigue or noisy pages, deciding what should page versus ticket, or when on-call is drowning in alerts nobody acts on.

Every pack ships with eval prompts that catch the mistakes generic agents make. A few of the 14 scenarios:

Checking access…