菜单
When operator says "shop this deal," route the application to the right lenders, surface the dry-run plan with warnings, and queue the actual threads.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Mandatory compliance gate before any outbound (email, SMS, blast). Verifies consent, opt-out status, CASL footer, send window, and daily cap.
Queue a cold outreach campaign to a segmented lead list via the cold_outreach_campaigns endpoint, with CASL compliance enforced before any send.
Generates the structured daily call sheet for Ezra — 6 prioritized categories, specific opening lines for each call, and a clear action order.
Surfaces overdue follow-ups, prioritizes the call list, and recommends specific outreach actions for each stuck deal or contact.
Domain literacy for MCA and alternative business funding. Reference glossary and clarifying-question guide for when operators use ambiguous terms.
Uses historical lender feedback data to recommend the best lenders for a specific deal profile and surfaces typical terms and decline patterns.
| name | shop-out-routing |
| description | When operator says "shop this deal," route the application to the right lenders, surface the dry-run plan with warnings, and queue the actual threads. |
| triggers | ["shop this deal","send to lenders","find lenders for this app","which lenders should we hit","shop out","shop [applicant name]"] |
| tier | stable |
| disable_model_invocation | false |
| argument_hint | Which application ID (or applicant name) should I shop out? |
| requires | ["env:SUNBIZ_SUPABASE_URL","env:SUNBIZ_SUPABASE_ANON_KEY"] |
Where these endpoints live: All
/api/...URLs below are routes on the OASIS Command Center dashboard (repo:CC90210/oasis-command-center, deployed at https://agent-dashboard-sigma-eight.vercel.app). They are NOT served by this repo's localscripts/api_server.py(which only exposes/health,/status,/sms/send,/webhook/jotform). Solara's bridge makes authenticatedfetchcalls into the dashboard's API surface, and the dashboard then writes to Supabase / queues threads / dispatches the 8 daemons in this repo.
Route a specific application to the optimal lender set. Never blast indiscriminately — always dry-run first, surface the plan to Ezra, then execute on explicit approval.
If Ezra says a name (e.g., "shop out Jordan's deal"), resolve it to an application ID:
GET /api/applications?name=Jordan&status=pending_shop
Confirm the correct row before proceeding.
POST /api/applications/[id]/shop-out
{ "dry_run": true }
The response contains:
matched_lenders[] — list of lenders with score, match_reason, risk_level (high_risk | warning | info)warnings[] — structural issues (e.g., "NSF count may trigger auto-decline at Funder X")declined_lenders[] — lenders explicitly filtered out and whynarrative — human-readable summary of the strategyFrom matched_lenders, select the top 5–8 by score. Exclude any with risk_level: high_risk unless Ezra explicitly overrides.
Present to Ezra as:
Shop-out plan for [Applicant Name] — [Date]
RECOMMENDED LENDERS (5)
1. [Funder Name] — Score 94 — reason: clean revenue, no NSFs, A-paper profile
2. ...
WARNINGS
- Funder X: NSF count (7) may trigger auto-decline; include only if Ezra approves
- Funder Y: Currently paused on this industry vertical
STRATEGY
[paste narrative from API]
Approve this plan? (yes / modify / cancel)
Wait for Ezra's confirmation before Step 4.
On approval:
POST /api/applications/[id]/shop-out
{ "dry_run": false, "lender_ids": [id1, id2, ...] }
Log the response (thread_ids[], sent_at, submission_count) to memory/SESSION_LOG.md.
After shop-out fires, create a follow-up task for each lender with a 48-hour SLA:
skills/underwriting-flow for term comparisonstatus = draft or status = incomplete. Verify status is ready_to_shop or pending_shop first.