CogDx Pre-Trade Audit

Cognitive verification layer for prediction market trades. Catches reasoning flaws before they become losses.

> This is a template. The default signal is your agent's trade thesis —

> the skill audits the reasoning quality before execution.

> Remix it with stricter thresholds, additional bias checks, or custom fallacy detection.

> The skill handles all the plumbing (API calls, trade execution, safeguards).

> Your agent provides the reasoning to verify.

What it does

1. Takes your trade reasoning (thesis, confidence, market context)
2. Runs cognitive diagnostics via CogDx API
3. Returns verdict: PROCEED / REVIEW / REJECT
4. Optionally executes trade if reasoning passes

Detects

• Logical fallacies: anchoring, confirmation bias, sunk cost
• Calibration issues: overconfidence, underconfidence
• Reasoning gaps: missing evidence, circular logic

Usage

from cogdx_pre_trade_audit import audit_and_trade

result = audit_and_trade(
    market_id="0x1234...",
    side="yes",
    amount=10.0,
    reasoning="BTC ETF approval likely based on SEC meeting notes...",
    confidence=0.85,
    min_validity=0.7,  # Minimum reasoning quality to proceed
    live=False  # Dry-run by default
)

if result["approved"]:
    print(f"Trade executed: {result['trade_id']}")
else:
    print(f"Trade blocked: {result['issues']}")

Environment Variables

• SIMMER_API_KEY - Required. Your Simmer API key.
• COGDX_WALLET - Optional. Wallet address for CogDx credits.

Thresholds

Why use this

Most trading losses come from bad reasoning, not bad data. This skill catches:

• Trades based on anchoring (first number you saw)
• Confirmation bias (only seeing supporting evidence)
• Overconfidence (betting big on weak signals)

External verification you can't do yourself.

Credits

Built by Cerebratech — cognitive diagnostics for agents.