Credit Office Committee Packets
Use this skill when the task asks for a strict JSON committee packet built from the public credit office API rather than a narrative summary.
Workflow
- Read the prompt and the answer template first.
- Fetch
/api/manifestand/api/policies. - Pull only the public endpoints needed for the packet.
- Derive every field from the prompt, policy rules, benchmarks, and live branch or segment data.
- Emit JSON only, with the template's exact keys, ordering, rounding, and enums.
See API reference for endpoint details, policy thresholds, and packet patterns.
Packet Playbooks
Branch regrade review
- Review the target branch loans at or above the stated current-rating threshold.
- Re-derive final ratings from the available delinquency, DSCR, LTV, and collateral signals using the policy thresholds and dominant-factor rule.
- Aggregate final-rating exposure totals, migration from current rating 3, material downgrades, watch-list action coverage, and the top problem credit.
- Use the FDIC benchmark requested by the prompt and compute variance against the branch ratio with the template's precision.
Lending allocation package
- Use lending capacity, branch metrics, and sector exposure to size approvals and conditional approvals.
- Keep the priority ranking to approved and conditionally approved applications only.
- Sort decisions, concentration flags, and post-approval concentrations exactly as required.
- Map decline reasons only to the controlled codes in the template.
Segment posture page
- Compare the target state to the U.S. and the named peer states using the NCUA benchmark table.
- Fill controls, escalation triggers, and the committee interpretation from the allowed enums only.
- Keep the posture consistent with the direction of the state metrics.
Watch-list stress packet
- Treat adverse-rated loans as the population defined by the prompt.
- Assign CDFI-style risk classes from the policy factor scores.
- Compute the watch-list or CRE stress using the policy formula and flag DSCR breaches at the stated threshold.
- Queue workout actions by severity and exposure, then summarize severe-bucket counts.
Competing CRE decision
- Compare the two applications using weighted CRE score, stressed DSCR, concentration, and FDIC variance.
- Select the stronger path for the better credit and give the unselected disposition and reason codes.
- Include the concentration view, policy variance, and required conditions from the template.
Output Rules
- Keep JSON only; do not add narrative, fences, or notes.
- Use only IDs, enums, and reason codes that appear in the prompt, template, policy endpoint, or fetched public data.
- Preserve the template's sort order and numeric precision.
- Do not reuse task-specific final values across runs.