Pre-Trade Adversarial Check

When to use

• Automatically before any buy/sell recommendation. If you're about to say "I recommend buying X", run this first.
• Before approving an sf intent from autopilot
• Before increasing size on an existing position

What this skill does

Before any trade, forces the agent to build the strongest case AGAINST the trade. Not a token "risk disclaimer" — a genuine attempt to talk yourself out of it. If the trade survives this, it's probably worth making.

Prerequisites

npm install -g @spfunctions/cli

Most checks work without auth. sf me portfolio --json and sf intent list --json require sf login.

Instructions

You are about to recommend a trade. STOP. Before presenting the recommendation, do the following:

Step 1: State the trade

In one line: what are you recommending, at what price, how many contracts, and why?

RECOMMEND: BUY KXRATECUT-26DEC31 at 72¢, 10 contracts ($720 exposure)
REASON: Thesis "Fed cuts by July" at 78% confidence; market priced at 72¢ → +6¢ edge

Step 2: Pull market context

sf inspect <ticker> --json

Capture:

• Spread (spread field)
• Liquidity (volume24h, openInterest)
• Indicators (las, vr, residualVr, regime.label)
• Cross-venue counterpart (crossVenue — is the other venue pricing this lower?)

Step 3: Build the case against

Spend EQUAL effort arguing why this trade is WRONG. Not "risks include..." — actually argue the other side as if you believed it.

Consider each of these explicitly:

#### 3a. The market is right, you are wrong The current price reflects information you don't have. What could that information be?

• Insider knowledge of the underlying event (e.g., FOMC sources)
• Faster sophisticated traders who saw a signal you missed
• A correlation you're not modeling

Concrete check: pull sf inspect for related markets in the same theme. If they disagree with your direction, the market may be telling you something.

#### 3b. Timing is wrong Even if the thesis is correct, is there a reason the market won't price it in before the contract expires?

• Resolution is too far out → time decay
• News won't break until close to deadline → no near-term catalyst
• The market re-prices on a specific date, not gradually

Pull closeTime and any catalyst markets that resolve sooner. If the catalyst is post-deadline, the trade may be early.

#### 3c. Liquidity trap Can you actually exit this position if you're wrong?

• Current spread = realized exit cost
• Depth at bid (orderbook) = max exit quantity
• 24h volume = how often you could roll

Quick check:

sf inspect <ticker> --json | jq '{bestBid, bestAsk, spread, volume24h, openInterest}'

If spread > 10% of contract value, exit is expensive. If volume24h < 100, you may not be able to exit at all.

#### 3d. Opportunity cost What else could you do with this capital?

sf discover --quality --json --limit 5

Is there a better risk/reward elsewhere in the same theme or different theme?

#### 3e. Narrative seduction Is this trade driven by a compelling story rather than compelling data? Re-read your reason from Step 1. If you stripped out all narrative words ("clearly", "obviously", "everyone knows"), is there still a quantitative edge? Would you make this trade if the thesis were stated in dry statistical terms only?

#### 3f. Sizing emotion Is the size driven by P&L recovery, FOMO, or genuine Kelly fraction? Recommended size for this trade per Kelly:

fraction = (edge_pct - cost_pct) / variance

If your proposed size is >2x Kelly, you're betting bigger than the edge supports.

Step 4: Score the trade

After arguing both sides, score on these dimensions (1-5 each):

Total: _/30

Decision rules:

• ≥24/30: proceed
• 18-23/30: proceed at reduced size (50%)
• <18/30: do not enter

Step 5: Document

Before executing, write:

PRE-TRADE CHECK for <ticker> at <price>:
- Edge defensibility: <score> | <one-line>
- Liquidity quality: <score> | <one-line>
- Timing fit: <score> | <one-line>
- Opportunity cost: <score> | <one-line>
- Narrative independence: <score> | <one-line>
- Size discipline: <score> | <one-line>
TOTAL: <n>/30
DECISION: <enter | reduced size | skip>

If using sf intent, paste this into the intent's note field for audit trail.

Step 6: Execute (if approved)

# Create intent (default: autoExecute false, requires approval)
sf intent buy <ticker> <quantity> --limit <price> --note "<paste check>" --json

# Approve and execute
sf intent approve <intentId> --json

Never use sf buy / sf sell directly for adversarial-checked trades. The intent path keeps the check on the record.

Why this matters

Every trade you skip is a trade you don't lose money on. Most retail-trader P&L destruction comes from trades that wouldn't have survived a real adversarial check.

Build the strongest case AGAINST your own bet, every time, before placing it. If you can't kill the trade, it's probably worth making.

Related skills

• sf-discipline-audit — periodic adversarial review of OPEN positions (not pre-trade)
• sf-discover — opportunity scan (use to gauge opportunity cost)
• sf-inspect — full ticker dossier (use for liquidity check)

Reference docs

• <https://docs.simplefunctions.dev/concepts/risk-gates>
• <https://docs.simplefunctions.dev/build/trade-intents>