ApexCloud Retention Ops
Use this skill for structured JSON deliverables built from the ApexCloud Retention Operations API.
Workflow
- Read the prompt and the provided answer template together.
- Treat the template as the schema contract. Copy its keys, nesting, and ordering exactly.
- Extract the report type, scope list, date range, month buckets, and as-of date before you query anything.
- Use the task-environment base URL from the prompt or workspace note, then query only the endpoint families the task actually needs.
- Fill the template with computed values, not placeholders. Choose one concrete enum value wherever the template offers a controlled vocabulary.
- Reconcile totals, rankings, and sort order against the detailed rows before you finish.
- Return JSON only.
Report families
When the prompt is one of the recurring ApexCloud report types, use the matching playbook in references/report-archetypes.md.
Output discipline
- Respect explicit scope limits. Do not expand beyond the listed account IDs or candidate IDs.
- Keep ranked lists in the order the prompt asks for.
- Use chronological month order unless the prompt says otherwise.
- Keep currency values to 2 decimals, percentages to 1 decimal, counts as integers, and probabilities to the requested precision.
- Preserve controlled labels exactly as written in the prompt or template.
- Use
nullonly when the template or prompt explicitly allows it. - Do not invent source systems, labels, or extra keys.
Sanity checks
- Does every required field exist in the final object?
- Do summary counts match the detailed lists?
- Are the selected labels valid members of the template vocabulary?
- Did you use the correct source family for each metric?
- Is the final response valid JSON with no markdown fencing or commentary?