The two decision engines — rule-based 7-Step and panel-based VOLTRAN — plus the full L1-L5 causal chain. Focused on a single question: what did the system decide, and why?
| STEP | GATE | FINDING | SCORE |
|---|---|---|---|
| S1 | Market Pulse | CFI EXPANDING, flow +0.46, crypto leading | 6.2/10 · PASS |
| S2 | Asset Scanner | SOL #3 of 47 scanned · composite 78/100 | 8.5/10 · PASS |
| S3 | Safety Check | Liquidity 81.3 · contract clean · no honeypot | 5.5/10 · PASS |
| S4 | Timing | MA alignment bullish · squeeze released · RSI 65 | 9.0/10 · PASS |
| S5 | Trade Plan | Entry $112.05 · SL $107.07 · TP1 $127.74 · R:R 3.7:1 | 9.5/10 · PASS |
| S6 | Trap Check | Bull trap @ $113.86 · layering · wash trading · liq grab | 2.0/10 · BLOCK |
| S7 | Final Verdict | Weighted S1-S6 aggregate · S6 veto triggers AVOID | 3.6/10 · AVOID |
Steps 1-5 scored strongly (avg 7.7/10) — macro, asset, safety, timing, and trade plan all clear. S6 blocks the pipeline: microstructure evidence (bull trap, layering, wash trading, liquidity grab at $4.5M cluster) overrides structural strength. The rule-based engine does not second-guess — trap detection is a hard veto.
| AGENT | REASONING | VOTE | WEIGHT | |
|---|---|---|---|---|
A |
ARIA TECHNICAL |
8 bullish signals · MA aligned · Elliott wave 3 · Ichimoku bull | LONG · 80% | 25% |
N |
NEXUS RISK |
R:R 3.69 · 3 warnings (layering, wash, liq) · 50% cap advised | MEDIUM | 25% |
O |
ORACLE ON-CHAIN |
Smart money LONG 85% · F&G 21 (contrarian) · SM divergence | LONG · 85% | 20% |
S |
SENTINEL SECURITY |
Bull trap $113.86 · layering · wash trading · $4.5M liq grab | ALERT · VETO | 30% + VETO |
Rule triggered: manipulation evidence is concurrent and confirmed across layering, wash trading, and trap signals. Directional consensus (3 LONG) is overridden by the veto rule.
Three agents vote LONG — technical, on-chain, and ML ensemble all favor the trade. NEXUS is cautious but operable at half size. SENTINEL's ALERT triggers the veto rule: concurrent manipulation evidence overrides directional consensus. If SENTINEL clears (bull trap invalidation), the panel's directional bias reverts to LONG.
How evidence propagated across layers to reach the final decision.
| TRANSITION | TRIGGER | DOWNSTREAM EFFECT |
|---|---|---|
| L1 → L2 | EXPANDING regime detected + flow score +0.46 | Asset scanner activated · SOL eligible for 7-Step |
| L2.S6 | Bull trap pattern @ $113.86 · layering detected | S6 blocks · 7-Step score 3.6/10 · AVOID verdict |
| L2 → L3 | 38 technical features forwarded despite AVOID | MLIS Pro engaged · independent ML second opinion |
| L3.K5 | Weighted composite: K1(94)+K2(60)+K3(75)+K4(60) = 73 | BUY verdict · direction AGREE with L2 · conviction DISAGREE |
| L3 → L4 | Dual-engine PARTIAL state · panel review requested | 4 agents activated for independent review |
| L4.SENTINEL | Layering + trap + wash trading concurrent evidence | ALERT raised · veto rule overrides 3-agent LONG consensus |
| L4 → L5 | Panel AVOID directive · override mode: 50% cap | Blueprint derivation with caution parameters |
| L5.GATES | G5 confidence 58% + G6 trap (inherited from L2.S6) | WAIT-state · no orders placed · blueprint held |
One microstructure observation — bull trap pattern at $113.86 with confirmed layering and wash trading — propagates through 4 layers: L2.S6 (primary detection) → L4.SENTINEL (security alert) → L5.G6 (gate fail by inheritance) → amplified throughout. This is not four independent opinions; it is a single observation cascading. The system's caution is one piece of evidence wearing four hats.
Two engines, one chain, one directive.
Rule-based pipeline: 5 steps pass strongly, S6 trap detection hard-blocks. Deterministic outcome — no interpretation required.
Panel-based synthesis: 3 agents LONG, SENTINEL ALERT triggers veto rule. Rule-based override of directional consensus.
Two engines, same conclusion, from different directions. The 7-Step detects the trap via pattern matching. VOLTRAN reaches the same conclusion via panel voting and security veto. The L1-L5 chain shows a single microstructure observation propagating through four layers — this is a reinforced decision, not a noisy one.