Implement E: commodity-wedge shadow check before first argument move (five binary axes)

Add commodityWedgeShadowCheck(hypothesisId) in argument.js alongside the describability shadow check, also gated on argument_moves_count===0. Single Gemini Flash call evaluating five binary axes: (1) incumbent_already_owns_workflow — an existing SaaS product (named) already covers this workflow for the ICP; (2) requires_deep_system_integration — first revenue requires CRM/ERP/procurement-grade API credentials from the buyer; (3) deliverable_is_dashboard_or_taxonomy — the product output is a report, classification, or analytics view rather than an outcome-linked action with external ground truth; (4) no_urgency_event_named — no compliance failure, quantified OPEX waste, or direct revenue leakage is named as the trigger in the hypothesis; (5) no_owned_data_advantage — the product would produce no cross-customer compounding dataset that a clone starting later could not replicate within 6 months. commodity_wedge_recommendation=true if ≥4 axes are true. Log to meta_engine/data/shadow_gate_verdicts/commodity-YYYY-MM-DD.jsonl. Expose as additional columns on the existing /commander/<token>/shadow_verdicts route. No kill behavior.

The 'RevOps Objection Taxonomy Normalizer' shape (GPT-5.5-Pro Round 1: passes describability, observed-buyer, solo-inbound, yet still structurally weak on urgency and data advantage) flags commodity_wedge=true on axes 3+4+5. After 4 weeks of shadow: hypotheses where commodity_wedge_recommendation=true should show higher council kill rate than the baseline council kill rate (~55% across all council-stage hypotheses per recent traces).

Before enabling shadow collection, create a 12-item calibration set (4 commodity wedges drawn from Round 1/2 red-team-named examples, 4 ROBUST non-wedges from the S157 5/5 list, 4 ambiguous). Shadow check must flag ≥3/4 commodity wedges, 0/4 robust non-wedges, and include ≥2 concrete axis names per flag. If calibration fails on any condition, shadow collection is blocked and prompt is revised. After 4 weeks: if ≥10 shadow verdicts exist and commodity_wedge=true does not predict council kill at a higher rate than random, axes are miscalibrated.

Rejected by filter_score. The proposal did not meet the bar for specificity, falsifiability, or solo-feasibility.