Responsible gaming

What the RG layer does, the hooks your platform can drive, and the invariant underneath all of it: every operator-facing control can make the feed stricter — none can make it looser.

The built-in baseline (non-configurable)

• Breather at 3 consecutive losses, global across games. The round finishes normally; the breather is the next card. Its continue button is gated for the full 8 seconds — no skip path. Session net (win or loss) is shown plainly.
• 2 non-bet cards guaranteed after every breather — betting-capable cards are hard-filtered from the candidate pool, not down-weighted.
• Loss-streak damping: at 2 consecutive losses the playable-card weight drops ×0.15 for the pick.
• RG modifications only ever subtract betting opportunity from the baseline mix, never add.

The one RG config field — clamped tighten-only

The clamp is unit-tested precisely on the weaken-attempt cases: a hostile or mistaken value can only tighten or be ignored.

Hook 1 · Pre-session player check

With wallet: 'operator', the feed posts fidra:v1:player_check on init. You answer with the player's self-exclusion / limit verdict:

{ source:'fidra-host', proto:'fidra:v1',
  type:'fidra:v1:player_check_result', allow: boolean, reason?: string }

• allow !== true → non-bet mode: every bet is blocked at the wallet boundary — no debit, no roundId, no bet_placed — and the feed keeps running watch-only.
• With rg.requirePlayerCheck: true, bets start disabled until an allow: true arrives — a self-excluded player can never get a bet in before the check returns.

Hook 2 · Forced pause

fidra:v1:pause inserts a breather now, reusing the exact state the 3-loss trigger sets. One-way lever: an operator can force a pause, never suppress one.

Hook 3 · Operator risk state

Your platform sees the whole account — deposits, cross-product losses, limits. Send a risk state at mount or mid-session (fidra:v1:risk_state { state }) and the composition tightens:

Every transition is written to the tamper-evident audit log with reason code operator_risk_state, and every composition decision stamps the risk state in force. Moving a player back toward normal is your clinical decision — fidra never loosens on its own.

What remains your duty

fidra's layer shapes presentation in response to a state you compute. Detecting harm (deposit velocity, chasing, self-exclusion, affordability), computing the risk state, and all statutory RG obligations remain entirely the operator's responsibility. fidra supplements your controls; it does not and cannot replace account-level harm detection.