prd

PRD — Star Trek: Expansion Capture & Intelligence Platform

Draft v0.1, July 2026 · Initiative owner: Gloria Amune, Product · Priority P0 · Confidential. Full source document: attachments/Star-Trek-Expansion-Capture-Intelligence-PRD-v0.1.pdf

This file covers the slice of the PRD relevant to this product (rechaj-star-trek-app, the Survey Agent mobile app). The Supervisor/Admin Web Dashboard lives under products/atlas/flows/star-trek-supervisor/ and star-trek-admin/; the Atlas console update is documented against the existing products/atlas/flows/star-trek/ shared module.

Problem

The original MVP put capture, cleanup, and reporting loosely inside tools built for other purposes (the Agent App, Atlas). The OCR plate-read pipeline isn’t trustworthy enough alone and needs a human review layer; pay is now tied to approved-image count, which needs a durable, auditable review trail the old passive duplicate-flag model didn’t have.

Goals & non-goals

Every image has an auditable trail: captured → reviewed → approved/rejected → (area) admin-approved.
Agents are paid accurately, only for Supervisor-confirmed-good work.
Out of scope for this app: actual payment disbursement (manual, outside the product), rider onboarding/recruitment (separate, unchanged), an appeals process for a rejected image (not built — the Supervisor’s call is final).

Audience

Survey Agent — captures photos of tricycles in their assigned area. Paid ₦250 per image a Supervisor approves. Not necessarily full MocX/Rechaj staff.

User stories

As a Survey Agent, I want to capture a photo of a tricycle quickly, so I can cover ground during my shift.
As a Survey Agent, I want to see how many of my photos have been approved and what I’ve earned, so I can track my own pay.
As a Survey Agent, I want to know which area I’m currently assigned to, so I know where to work.

Functional requirements (Star Trek App, agent-facing)

Home screen: current assigned area (read-only), one clear Capture action.
Capture: one photo required; GPS + timestamp + device + agent ID captured automatically; optional notes field.
Two-stage plate detection (on-device at capture, backend after upload) — carried over from the original design, purely informational to the agent, never blocks submission.
Own progress view: images captured today, approved count, running pay total (filterable by date range).
Requires an internet connection to capture — unlike the original Agent App design, this is not offline-first.
Optional notes field, and the Capture/Success screen flow, carry over unchanged from the original design.
Payment ledger (agent’s own), filterable by date range and status — visible here, in the Web Dashboard (Supervisor/Admin, scoped to their people), and in Atlas (Leadership reporting).

Core flow touchpoints (agent’s view of the area lifecycle)

Agent is assigned to exactly one active area at a time by their Supervisor.
Agent captures continuously; each image is Pending until a Supervisor reviews it, then Approved or Rejected.
If the Agent is blocked mid-area, their area stays open for a new Agent — but the blocked Agent’s own already-approved images remain valid for their own pay once the area is eventually approved. A blocked Agent immediately loses app access.
Pay is not calculated per image in real time — it’s finalized only once Admin approves the whole area, so a Supervisor’s in-progress approvals are provisional until then.

Data model touchpoints (Observation / Image)

Observation ID, Photo, GPS Coordinates, Timestamp, Agent ID, Device, Area ID, Notes, plate_status (Detected / Undetected), review_status (Pending / Reviewed), cleanup_status (Approved / Rejected, once reviewed), is_potential_duplicate (True / False / Unresolved — automated signal alongside the Supervisor’s manual review).

Open questions

Exact review turnaround expectations — is there a target for how quickly a Supervisor should clear their pending queue?
Full field-by-field data model, to be finalized with Engineering.