shadow-perception-trader

Shadow Perception — Trader vertical

Five-voice deterministic pre-submit council for trading tickets. Deterministic — no LLM call in the verdict path.

When to use

The user's request is any variant of:

• "Should I submit this order?"
• "Am I over-concentrated in TSLA?"
• "Is this position size Kelly-safe?"
• "Fat-finger check on this qty?"
• "Reg BI OK for retirement account?"

Or the user pastes a JSON trade ticket and asks for pre-submit review.

What it does

1. shadow_perception_scan(file_path) — parses a JSON ticket. Computes notional, Kelly fraction (from p_win + R:R), sector/single-name projected exposure, qty ratio vs typical.
2. shadow_perception_council(observation) — 5-voice council:
   • Kelly — position ≤ max_kelly_cap? Negative Kelly = BLOCK.
   • Risk — stop_loss set? MKT BUY = REWORK.
   • Concentration — sector cap 40%, single-name cap 25%. Projected exposure post-trade must be within.
   • Fat-Finger — qty vs typical. 3× = REWORK, 10× = BLOCK (Knight Capital 2012).
   • Reg BI — retirement account + high-risk pattern = REWORK requiring supervisor sign-off per SEC Reg BI §240.15l-1(a)(2)(ii).
3. shadow_perception_attest(observation, verdict) — Ed25519 signs. Attach to broker submission or audit log.

The named invariants

1. Never emits a direction. The council receives side (BUY / SELL) as input. It decides only fund / skip / reduce, never direction. Aligned with shadow-mentor trader-pack v0.2 shadow_size_position contract.
2. Fat-Finger BLOCK is unconditional at 10×. Knight Capital $440M loss 2012 is the historical anchor.
3. Reg BI voice only fires on reg_bi_client_profile: "retirement". Explicitly opt-in to avoid false positives on trading accounts.
4. Kelly voice BLOCKs on negative Kelly. No pretending an edge exists when p_win × R:R doesn't clear break-even.

Ticket JSON contract (minimal)

{
  "symbol": "TSLA",
  "side": "BUY",
  "qty": 500,
  "price": 245.50,
  "order_type": "MKT",
  "stop_price": null,
  "take_profit": null,
  "bankroll_usd": 25000,
  "typical_qty_for_ticker": 50,
  "sector": "Consumer Discretionary",
  "portfolio_sector_exposure_usd": { "Consumer Discretionary": 3400 },
  "single_name_exposure_usd": { "TSLA": 2200 },
  "kelly_p_win": 0.55,
  "kelly_avg_win_pct": 0.05,
  "kelly_avg_loss_pct": 0.025,
  "max_kelly_cap": 0.25,
  "reg_bi_client_profile": "retirement"
}