Skip to content

TRACK_ANSWERS.md

Latest commit

 

History

History
122 lines (77 loc) · 4.5 KB

TRACK_ANSWERS.md

File metadata and controls

122 lines (77 loc) · 4.5 KB
Raw

Track Answers — Build On Stellar / IBW 2026

Sub Rosa won 1st Place in the Hack Privacy Track. The multi-agent/x402, mainnet smoke, SDK, and ecosystem paths stay in the product as supporting evidence that the privacy primitive can become reusable Stellar infrastructure.

Summary

Submission area Strength Weak point (before) Addressed by
Hack Privacy tlock + on-chain BLS in code Separate threat model + auditor UI missing docs/THREAT_MODEL.md, Auditor tab + e2e blob decrypt
Infrastructure support Reusable primitive, SDK, mainnet, ecosystem depth OZ Relayer / Passkey "scored flex" thin OZ Channels adapter, Passkey-Kit demo, docs/ECOSYSTEM.md
Supporting agent proof Multi-agent live e2e, mandate, x402, cap negatives Cap partly off-chain Passkey tab, honest cap copy, docs/THREAT_MODEL.md

Infrastructure support

What we prove

  • Reusable Round Soroban contract with full lifecycle
  • SDK with spec-accurate bindings from wasm
  • Mainnet deploy + settlement (CA7KSDEY…)
  • Watch-mode keeper for liveness
  • Ecosystem hooks documented (Passkey-Kit, Smart Account Kit, Scaffold Stellar)

OZ Relayer (scored flex)

Status: Optional submitter adapter implemented; direct RPC remains default.

packages/sdk accepts a submitter, and createOzChannelsSubmitter() sends signed XDR through OpenZeppelin Relayer Channels. All existing e2e scripts still use direct Soroban RPC by default, preserving the proven path.

Passkey-Kit / Scaffold Stellar (scored flex)

Status: Passkey-Kit is installed and lazy-loaded in the UI wallet panel.

The panel creates a real browser passkey via passkey-kit (embedded testnet WASM). Optional Deploy smart wallet may fail without sponsor funding — not required for submission. Agent mandate enforcement intentionally stays on Ed25519 session keys in this build.

Privacy track

What we prove

  • Drand tlock seal with live quicknet
  • On-chain BLS12-381 verification at open_reveal
  • Auditor blob (X25519 sealed box) in packages/tlock
  • Seal attack demo (seal-off vs seal-on) in UI

Threat model

Status: docs/THREAT_MODEL.md — adversaries, mitigations, residual risks, honest off-chain limits.

Auditor UI

Status: apps/web Auditor tab:

  • Paste auditor secret (hex) → decrypt identity blobs emitted by pnpm agents:e2e
  • Live bid decrypt against published R via quicknet

Drand R countdown

Status: Live quicknet countdown in site header + lifecycle banner.

Supporting agent proof

What we prove

  • Two autonomous agents (pnpm agents:e2e) — exceeds PRD single-agent minimum
  • Signed session mandate with maxBid / maxEscrow / maxAppraisalSpend
  • x402 appraisal with on-chain USDC settle
  • 10 unit tests including cap negative scenarios
  • Agent commits via session key, not principal

Passkey wallet policy

Status: Documented in Passkey tab + docs/ECOSYSTEM.md. Mandate caps map to Smart Account context rules / spending policies in production.

Cap enforcement honesty

Cap Layer Enforced?
maxBid Agent (mandate) Off-chain only
maxAppraisalSpend Agent (mandate) Off-chain only
maxEscrow Agent + commit tx Off-chain + public escrow amount
bid ≤ escrow Round contract On-chain at reveal

UI Caps tab and CAP_SAFETY_COPY state this explicitly.

Winner settlement vs x402

Status: SettlementRail panel + docs/TECH_DESIGN.md.

  • x402 = appraisal micro-payment (agent → server)
  • settle() = winner prize (escrow → operator via SAC)

Same USDC asset rail (SEP-41); different authorization. Winner payment is not x402 by design — contract-enforced escrow settlement.

Intentional PRD deviation

  • Multi-agent (2 bidders) instead of PRD's single autonomous agent — stronger supporting proof for the Hack Privacy winning primitive.

Mainnet vs testnet (submission)

Testnet Mainnet
Full USDC lifecycle ✅ lifecycle:e2e ❌ not in scope
Multi-agent + x402 + settle ✅ agents:e2e (one canonical trace) ❌
Auditor UI blobs ✅ from agents:e2e → demo-trace.generated.ts ❌
Deploy + BLS + settle ✅ ✅ 1/5 XLM smoke
Read-only verify — ✅ pnpm mainnet:verify

Mainnet intentionally uses micro XLM amounts, never testnet demo sizes (700/459 USDC).

Demo script

See docs/DEMO_SCRIPT.md for jury walkthrough order.