HarborCRM Data Workflows
Use this skill when the task is to turn HarborCRM API data into an exact JSON deliverable.
Start from the template
- Read
input/payloads/answer_template.jsonfirst. - Treat required keys, enums, ordering, and nullability as the contract.
- Do not add fields.
- If the template shape changes, follow the template rather than reusing a previous answer shape.
Identify the family
- If the template has
sponsor_statuses,qualified_lead_accounts,follow_up, orcrm_action_counts, use the sponsor-handoff workflow. - If it has
qualified_exhibitors/excluded_near_misses/aggregate_countsorranked_leads/excluded_exhibitors, use the trade-show prospecting workflow. - If it has
clean_contacts/duplicate_summary/removal_summary, use the import-cleanup workflow. - See references/patterns.md for the exact rules.
Query only what you need
- Use the runner-supplied base URL from the prompt or environment note.
- Read
/api/policieswhenever the prompt mentions controlled enums, platform coverage, or qualification rules. - Pull CRM records before deciding create vs update vs exclude.
Normalize before comparing
- Emails: trim and lowercase.
- Phones: strip everything except digits.
- Dates: keep ISO
YYYY-MM-DD. - Money and counts: integer USD / integer counts.
- Preserve
nullonly when the template allows it; otherwise use the template’s empty-string convention.
Finish cleanly
- Sort exactly as the template asks.
- Emit JSON only.
- Recheck that every count is derived from the records you actually included or excluded.