Safety · Maya anti-cheat moderation

Purpose#

Read-only moderation dashboard for competitive-session reports, appeals, tribunal votes, and active bans. It projects a deterministic MayaAnticheatModerationSystem snapshot so reviewers can inspect policy shape and evidence density without implying that the seeded rows are live cases.

Entry points#

• Direct route /trust-safety/maya-anticheat.
• Missing session redirects with reason=missing-session; a session without Trust & Safety workspace access redirects with reason=forbidden-workspace.
• Renders inside AdminShell with currentWorkspaceId="trust-safety".

Layout regions#

• Header: Maya Nexus identity, concise purpose, generated time, and explicit “seeded control snapshot” source label.
• KPIs: active bans, priority report targets, open tribunal cases, and pending appeals.
• Report aggregation: suspected target, independent reporters, category mix, evidence ids, and priority-review state.
• Appeals: appeal/ban identity, scope, reason, state, and reviewer resolution.
• Tribunal: required vote progress, per-reviewer verdicts, and final outcome.
• Active bans: player, scope, reason, issue/expiry times, and evidence.

States#

• Authorized — exact seeded dashboard counts and all four evidence sections render.
• Missing session — protected data does not render before redirect.
• Forbidden workspace — sessions outside the Trust & Safety scopes are denied before dashboard creation.
• Priority aggregation — independent-report threshold is exposed in both row data and visual treatment.
• Pending / resolved appeal — reviewer identity appears only when a resolution exists.
• Open / decided tribunal — vote count and verdict remain explicit.
• Mobile — all tables scroll within their wrappers while the page itself remains overflow-free.

Interactions#

• The dashboard is intentionally read-only; no row or button suggests that a reviewer can issue a ban, resolve an appeal, or cast a vote here.
• Table and list traversal announces headers, verdicts, reviewer identity, dates, and evidence in a coherent reading order.

Data & contracts#

• Reads: createSeedMayaAnticheatDashboardData, which instantiates MayaAnticheatModerationSystem from @maya/anticheat, submits seeded reports, registers reviewers, opens tribunals, casts votes, issues bans, and creates appeals before taking the snapshot.
• Writes / realtime / persistence: none.
• Injection seam: bindMayaAnticheatDashboardLoader can replace the seeded loader in-process; no production binding is visible in the route.

Cross-references#

• Dashboard: apps/oshun/admin/src/app/trust-safety/maya-anticheat/MayaAnticheatModerationDashboard.tsx
• Seed projection: apps/oshun/admin/src/app/trust-safety/maya-anticheat/data.ts
• Unit coverage: apps/oshun/admin/src/app/trust-safety/maya-anticheat/maya-anticheat-dashboard.test.tsx
• Browser coverage: apps/oshun/admin/e2e/maya-anticheat-moderation-dashboard.spec.ts

Open questions / known gaps#

• Replace the module-level loader seam with a tenant-aware BFF read and explicit loading, empty, unavailable, and stale states.
• Keep adjudication mutations on a separately scoped, rationale-gated workflow with immutable audit evidence rather than adding casual row actions to this overview.