Private Wealth Advisory JSON
Workflow
- Read the prompt,
input/payloads/request_memo.md, andinput/payloads/answer_template.json. - Identify
client_id,task_id, requestedanalysis_type, and any planning horizon in the memo. - Query the advisory API with
API_BASEfor:/api/clients/{client_id}/api/source-documents/api/retirement-accounts/api/life-insurance/api/trust-candidates/api/policies/tax/api/rmd-factors
- Filter broad endpoint responses to the requested
client_id; do not use records for other clients in the answer. - Resolve conflicts by source quality, not recency alone:
SIGNED_PROFILEcontrols profile, goals, beneficiary count, and tax rate when present;CUSTODIAN_EXPORTcontrols retirement balances; life-insurance planning records are reported as controlled bySIGNED_PROFILE; trust-candidate assets are reported as controlled byATTORNEY_MEMO. - Calculate with the formulas in references/advisory-formulas.md, round USD values to two decimals, emit numbers as JSON numbers, and return only the final JSON object.
Helper Script
For these task families, prefer the bundled deterministic helper:
python skill/scripts/solve_advisory_task.py input --api-base "$API_BASE"
Use --task-id only if the script cannot infer the task folder name from the input path. Inspect the generated JSON before returning it, especially that it conforms to the provided template and any sorted-list requirement.
Manual Notes
- If calculating manually, read references/advisory-formulas.md first.
- Copy the template's required top-level keys exactly; include additional computed fields only when they are part of the learned advisory pattern and fit the requested object.
- For
action_set, sort enum strings alphabetically.