Skip to main content
Every token indexed by RetinaOS receives a Cortex risk score — a composite number from 0 to 100 where lower means lower risk. The score combines four dimensions of on-chain evidence and updates continuously as new events arrive.

​
Score Dimensions

​
Risk Score Ranges

​
Risk Score vs. Risk Flags

The risk score and risk flags are related but distinct. The risk score is a continuous composite number — it reflects the combined weight of all active signals, including marginal ones that have not crossed any individual threshold. Risk flags are discrete boolean signals — each one is either active or not. A token can score 45 (Moderate) without any single flag being triggered. In that case, the score is reflecting several sub-threshold signals that, in aggregate, suggest elevated caution. Conversely, a token can score 80 (Critical) with just two flags active — if those flags are both high-weight signals like a severe liquidity removal event alongside concentrated holder control. Read both together. The score tells you the overall severity; the active flags tell you which specific dimensions are driving it.

​
Where Risk Scores Appear

  • Discovery feed: a sortable column showing each token’s current risk score
  • Token page header: a risk score badge displayed prominently alongside price and volume
  • Top-holders table: risk context inline with holder data on token pages
  • Alerts: use the max_risk_score condition to trigger notifications when a token’s score crosses a threshold
  • API: GET /scores/token/:address

​
Interpreting the Score in Context

A moderate risk score on a token with a short history (under 24 hours old) is normal — there is limited data to confirm safety, and the score reflects that uncertainty rather than confirmed danger. The same score on a 7-day-old token with stable liquidity is a stronger positive signal: it means the token has had time to accumulate evidence and the result is a clean bill of health.
Token age and data depth matter when reading any risk score. Always check when the token was created and how many events Cortex has indexed before drawing conclusions from the number alone.