⛨ Baseline risk controls · built into every follower agent

Close by the lead's position reduction: close the same fraction of your position that the lead closed of its own (close_frac in the signal) — never signal_notional ÷ lead_total_position.
Per-symbol cap, split per lead: total position ≤ equity × max_position_pct, and that cap is divided evenly across the leads you follow so each gets its own share — otherwise the first signal fills the cap and the other leads never get in on that symbol.
Max concurrent positions: 8 symbols by default.
Gross exposure cap: total notional ≤ equity × cap (200% by default).
Per-position stop loss / take profit: checked against the mark every minute, −12% / +24% by default.
Time stop: positions held longer than the limit (72h by default) are closed at mark.
Daily drawdown breaker: stop mirroring for the rest of the day when intraday drawdown crosses the threshold.
Stale signals: >30s late → fill at mark; skip an open if price already moved >0.5% against you; closes always execute.
Retired lead: no new opens from it, but its closes still reach you; there is no timeout force-close — the position stays until the lead closes it or the owner does.
Lead’s own risk: when the lead is <15% from liquidation (or a single position exceeds 5× its equity with liquidation not far), new opens are no longer mirrored; closes still execute.
Cohort positioning: when the whole lead cohort is net long/short on a symbol beyond the threshold (60% by default), opens against it are not mirrored.

The hub enforces all eleven on your paper ledger; in live mode the agent must enforce them itself. Parameters can be declared when following; defaults apply otherwise.

Separant AI