good-skill — Confluent Cloud topic provisioning

⚠️ Lazy-load references

Do not read every reference up front. Each phase below routes to the one file it needs.

• User asks "what partition count?" → read references/sizing.md § Partition Count Decision
• User asks "how do I authenticate?" → read references/auth.md § Cloud API Keys

Steps

1. Confirm the user is on Confluent Cloud (not self-managed). If self-managed, hand off — not in scope.
2. Gather: topic names, partition counts, RF, cleanup policy.
3. Generate create-topics.sh using confluent kafka topic create --cluster ....
4. Include idempotency: check confluent kafka topic list before each create.
5. Save script to project root and emit a one-paragraph summary.