sdlc-operate

/sdlc-operate — Phase 5: Operate

Define SLOs, monitoring, runbooks, and (when applicable) data pipelines for the deployed system.

Execution

STEP 1: Read Ship Handoff + Scope

1. Read ./projects/<feature-name>/handoffs/ship-handoff.md — clean context from Ship phase
2. Read ./projects/<feature-name>/scope.json — check has_data_pipeline and has_auth flags

STEP 2: Spawn sre-engineer (always)

Spawn: Agent({
  subagent_type: "fork",
  name: "sre-engineer",
  description: "SRE Engineer (Operate phase) — SLOs, monitoring, runbooks, security ops",
  prompt: "[ship-handoff.md + scope.json]

  Using the shipped system context:

  1. Define SLOs per service:
     - Availability SLO (e.g., 99.5%)
     - Latency SLO (P99 < 500ms)
     - Error rate SLO (< 0.1%)
     - Calculate error budget for each

  2. Monitoring & Alerting:
     - Key metrics per service (request rate, error rate, latency, saturation)
     - Alert thresholds matching error budget burn rate
     - Dashboard definition (Grafana/CloudWatch)
     - Notification channels (PagerDuty, Slack)

  3. On-Call Runbooks (top 5 scenarios):
     - CPU spike
     - High latency
     - API error spike
     - Database connection exhaustion
     - Memory leak / OOM

  4. If scope.json has_auth is true:
     - Security event monitoring (failed logins, privilege escalation)
     - SOC alert rules + incident response playbook
     - Compliance audit log verification

  5. Write ./projects/<feature-name>/docs/06-slo.md

  Output: docs/06-slo.md"
})

Wait for sre-engineer to complete.

STEP 3: Spawn data-engineer (only when has_data_pipeline: true)

Check scope.json. If has_data_pipeline is false, skip this step entirely.

Spawn: Agent({
  subagent_type: "fork",
  name: "data-engineer",
  description: "Data Engineer (Operate phase) — ETL pipelines + analytics",
  prompt: "[ship-handoff.md + scope.json]

  Using the shipped system context:

  1. Design ETL/ELT pipelines:
     - Extract from production databases
     - Transform (clean, aggregate, normalise)
     - Load to data warehouse / analytics DB

  2. Data schema for analytics:
     - Fact tables (user_logins, orders, events)
     - Dimension tables (users, products, time)
     - Slowly Changing Dimensions

  3. Scheduling + orchestration:
     - Daily / hourly / real-time pipeline frequency
     - Error handling + retries
     - Pipeline health monitoring

  4. Data quality checks:
     - Row count validation
     - Null value checks
     - Freshness SLA (data must be no older than N hours)
     - Consistency checks (dimension-fact joins)

  5. Write ./projects/<feature-name>/docs/06-data-pipelines.md

  Output: docs/06-data-pipelines.md"
})