hyperbuild-15-adversarial-reviewlisted

# Step 15 — Adversarial review (parallel, 3 critics) You are executing step 15 (adversarial-review) of the hyperbuild pipeline. Step 14 implemented every epic with per-epic reviews and green tests; step 16 (ship gate) runs next and its verdict is final — this step is the last chance to find and fix defects before the gate. **Goal:** three independent adversarial passes over the WHOLE app — code quality, spec coverage, mockup fidelity — merged into one ranked findings list, applied as surgical patches, with structural gaps routed back through step 14 exactly once at most. **Loop gate (check FIRST):** if `runs/<run_tag>/gates/review-loop-log.md` exists, this is the SECOND pass — the one permitted loop already ran. On the second pass you re-run the full procedure below but you MUST NOT create new task files: remaining structural findings are recorded as known gaps (procedure item 10). If the loop log does not exist, this is the first pass. --- ## Inputs Read these before spawning anything: - `runs/<run_tag>/manifest.json` — run_tag, gear, platform, design_choice, stage (must be BUILD). If the router did not name the active run_tag, take the run whose manifest has `stage: BUILD`. - `runs/<run_tag>/idea.md` — verbatim app idea. GOSPEL. Paste into every spawn. - `research/product-spec.md` — the PRD: MoSCoW feature list, personas, screen inventory (step 4). - `research/02-engineering/author/stack-guide.md` — committed stack decisions (step 5). - `runs/<run_tag>/scaffold.md` —