Review
Decisions the system needs from you. Every flow surfaces its evidence β the search string, the source URL, or the breadcrumbs of reasoning β so you can validate honestly.
Contact Enrichment Review
0 pendingHunter posts enriched contacts (phone/email/LinkedIn) for human spot-check before they hit Salesfinity. Surfaces source URL + waterfall trail.
Exa Search Recipe Composer
0 pendingCompose a custom Exa question set for a discovery+enrichment run. Operator picks from engine/lib/exa_questions.json catalog; engineer-approved Qs only. Approved recipes drive ExaSkill.refresh() / discover_deep().
HVAC Warm-up Snippet Review
250 pendingValidate Exa-generated personalization (acquisition status, prior exits, founder story, awards, news) before pushing HVAC owner contacts into Salesfinity for a Next Chapter sell-side pitch.
Letter Review
102 pendingReview personalized M&A outreach letters before physical mail send via Stannp. Score-gated: β₯85% auto-approve suggestion, 65-84% QA review, <65% needs work.
Phoenix Lunch-and-Learn β Home Services Vertical Fit
0 pendingBear and Charlie post draft Exa Websets queries with cost estimate and per-vertical owner-fit reasoning. Ewing or Mark approves, rejects, requests revision, or approves with a forced cost cap. Approval triggers the server-side exa.websets.create call.
Phoenix Lunch-and-Learn β Vertical Fit
7 pendingCharlie scores Phoenix-area companies for the lunch-and-learn target list. Human approves/rejects each based on Google Maps source, website evidence, and mission-brief reasoning.
Propagate template change
0 pendingOperator approves applying a voice_prompt (or other authoritative field) change from one template to its hot/warm sibling. Source: mission #030.
System Notices
0 pendingInternal/demo flow for system-generated welcome and infrastructure notices. Not for agent output.
Generic Tasks (legacy)
2 pendingCatch-all bucket for the original review_queue. Anything posted without a flow_key lands here. Evidence not required (back-compat).
Transcript Correction
0 pendingReview fuzzy-matched transcript fixes proposed by the polish bot. Auto-promote runs at N=3 exact-agreement; this queue only surfaces when 3+ proposals fuzzy-agree (e.g. "DeChant" vs "Dechant") and a human pick is required.
Wrong-contact replacements
0 pendingOperator approves the same-title replacement person/phone after the rep flags a wrong contact or bad number. Source: mission #027.
Call Review
0 need actionAll calls scored by revenue opportunity. Approve next steps, edit drafts, or let agents handle follow-up.
Contact Enrichment
0 fields pendingPer-fact review of Exa-purchased contact data β distinct from the UVD task queue.
postReview() from @/lib/uvd. Add a new flow with the uvd-flow skill.