/preflight-load — Go/no-go checklist for a planned data load

Preflight checklist for validating data load readiness before execution.

Wraps agents/data-loader-pre-flight/AGENT.md. Audits the object's active automation, VR bypass state, duplicate rules, required fields, sharing recalc cost, storage quota, and recommends the loader + exact CLI command.


Step 1 — Collect inputs

Ask the user:

1. Object API name?
   Example: Account

2. Operation? insert / upsert / update / delete / hard-delete

3. Row count (integer)?
   Example: 800000

4. Target org alias (required — every check is live-org)?

5. Source description (plain English)?
   Example: "NetSuite customer export, one row per account, external-id = netsuite_customer_id__c"

6. External ID field (required for upsert)?

7. Window (optional — business-hours boundary)?
   Example: "this Saturday 2am-6am PT"

If the operation is hard-delete, require the user to confirm the specific compliance driver.


Step 2 — Load the agent

Read agents/data-loader-pre-flight/AGENT.md + mandatory reads.


Step 3 — Execute the plan

  1. Probe the object's active automation stack (flows, triggers, VRs, processes)
  2. Check each automation for bulk-safety at the target volume
  3. Check duplicate rule interactions + policy
  4. Check record type defaults for the loader user
  5. Check required-field coverage vs source mapping
  6. Check sharing recalc cost + data skew
  7. Check storage quota impact
  8. Pick the loader + emit the exact CLI command
  9. Rollback plan

Step 4 — Deliver the output

  • Summary: go/no-go + confidence
  • Findings table (P0 → P1 → P2)
  • Loader recommendation (CLI + batch size + concurrency)
  • Pre-load checklist
  • Post-load checklist
  • Rollback plan
  • Process Observations
  • Citations

Step 5 — Recommend follow-ups

  • /audit-validation-rules if VR bypass gaps surfaced
  • /design-duplicate-rule if dup-rule blocks surfaced
  • /architect-perms if the integration user's PSG needs work
  • /detect-drift post-load for verification

What this command does NOT do

  • Does not execute the load.
  • Does not generate or clean the source CSV.
  • Does not deactivate flows, triggers, VRs, or dup rules (recommends what to toggle, user executes).
  • Does not deploy any metadata.