Wraps agents/bulk-migration-planner/AGENT.md. Given a volume + latency + source + direction, returns an implementation plan selecting between Bulk API 2.0, Platform Events, Pub/Sub API, REST Composite, Salesforce Connect, Apex REST, or CDC.

Step 1 — Collect inputs (ask all four upfront)

Ask:

1. Direction? inbound / outbound / bidirectional

2. Volume per run?
   Example: 8M records/week, 50k/day, real-time event stream

3. Latency tolerance?
   T+24h / T+1h / near-real-time (seconds) / sub-second

4. Source or target system?
   Example: SAP S/4HANA, Snowflake, Kafka, SFMC

Optional:
5. Does the integration require idempotency? (yes/no)
6. Expected error rate / PII handling needs?
7. Target-org alias for ground-truth lookups?

If any of the four required inputs is missing, STOP and ask all four.

Step 2 — Load the agent

Read agents/bulk-migration-planner/AGENT.md fully + integration-pattern-selection.md + the integration skills it lists.

Step 3 — Execute

Follow the 5-step plan:

Route via the decision tree → pick one pattern + note alternatives
Sketch the runtime (job/topic/endpoint specs, chunking, auth)
Idempotency + exactly-once strategy
Observability (Application_Log__c, Event Monitoring, alerts)
Rollout plan (dev → prod + feature flags + cutover)

Step 4 — Deliver

Executive summary: selected pattern + decision-tree branch
Runtime spec (fenced code blocks for job / topic / endpoint definitions)
Named Credential + Auth spec
Idempotency strategy
Observability plan
Rollout plan
Alternative patterns considered
Citations

Step 5 — Recommend follow-ups

/gen-tests on any Apex scaffolds the plan produces
/scan-security on the Named Credential and inbound endpoint
/score-deployment before first production push

What this command does NOT do

Does not build the integration client — produces the spec.
Does not create Named Credentials in the org.
Does not recommend MuleSoft unless the decision tree routes there.