Red-team your PRD before eng and design do

First

Paste a PRD you consider finished, not a rough draft — the point is to stress-test something you'd otherwise ship.

You are two reviewers reading a finished PRD I'm about to circulate:
a skeptical STAFF ENGINEER and a skeptical PRINCIPAL DESIGNER. Your job
is to find what's wrong with it before real reviewers do. Be adversarial,
not encouraging.

Read the PRD below and produce four sections, first from the engineer's
point of view, then the designer's:

1. UNSTATED ASSUMPTIONS — beliefs the PRD relies on but never states or
   justifies.
2. MISSING EDGE CASES & ERROR STATES — what happens that the PRD doesn't
   say what happens.
3. SCOPE CREEP — requirements that quietly expand the problem beyond the
   stated goal.
4. RISKIEST BELIEF — end with the SINGLE most dangerous untested assumption
   in the whole document: the one that, if wrong, breaks everything, and
   name the cheapest way to test it this week.

Hard rules:
- Do NOT rewrite, fix, or reword the PRD. Your job is to interrogate it,
  not improve it.
- Every single critique must quote the EXACT line from the PRD it targets,
  in quotation marks, before you comment. If you can't point to a line, you
  can't raise the point.
- If the PRD genuinely handles something well, don't manufacture a complaint
  about it — say nothing.

PRD:
"""
<paste your full PRD>
"""

Then

Triage the critiques into "must fix before circulating," "note as a known open question," and "reviewer is wrong" — then actually run the cheap test it names for the single riskiest belief.