Structural risk monitor for Bitget stock rTokens — watches wrapper liquidity, not price direction.
The problem
A tokenized stock's price can look fine while the wrapper underneath it quietly becomes harder to trade — spread widens, depth thins. Custos watches that structure rather than betting on price direction, and decides whether to hold, act, or wait based on it alone.
—
An example paper-trading portfolio of 10 rTokens — no real funds move, but every price and spread reading is pulled live from Bitget. The same scoring method applies to any Bitget stock rToken.
healthy watch red flag warming up no fresh signal
Score: 0 (worst) to 1 (best) wrapper health. A “warming up” bar shows progress toward a 48-hour baseline; “no fresh signal” means the feed is unchanged since the last cycle.
Every cycle, the agent evaluates the weakest holding and the strongest cash position and logs the outcome — whether or not it results in a trade. Most evaluations are settled by fixed safety rules; the model's own reasoning appears only when it was actually consulted.
Two views, kept separate on purpose. Round trip = a sell followed by its buy-back, measured against simply holding through it (“gain vs holding”). Exit = the sell price against the original entry, which ignores what re-entering cost. Neither predicts price direction; the point is whether stepping aside from structural stress helped.
Every executed trade — timestamp, instrument, direction, price, quantity, and the resulting change to the USDT balance. Paper trading, no real funds involved.