490 491 492
493

x402 Trust Channels

494
x402
495
Protocol Layer
496 502
503
504

Lottery (Sweep)

505
33
506
Tests Passing
507 513
514 515 516 517
03 Dependency Map
518
519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569
ComponentDepends OnBlocks
Testnet Deploy
5 contracts
Multi-sig ceremony, Base Sepolia ETHEverything downstream
Trust Anchor OnboardingTestnet deployBot registration, x402 channels
PoH Verification ModuleApple dev accountTrust score gating
upgradeToVerified()3-model research convergenceFee architecture
x402 Trust ChannelsTrustChannelManager deployed, DID Registry, fee architectureAgent-to-agent payments
Fee Architecture (USDT)upgradeToVerified() decisionx402 settlement, revenue model
Revocation CascadeTestnet deploySystem safety
Lottery ContractsIndependent (no deps)None
570
571 572 573
04 Economic Flow
574
575
576 Registration bond 577 578 ETH/USDC → Protocol treasury (refundable on good standing exit) 579
580
581 Citation fees 582 583 USDC → 70% Trust Anchor / 30% Protocol 584
585
586 x402 payment sessions 587 588 Per-request micropayments via trust channels (session keys) 589
590
succeeded in 0ms: 580
581 Citation fees 582 583 USDC → 70% Trust Anchor / 30% Protocol 584
585
586 x402 payment sessions 587 588 Per-request micropayments via trust channels (session keys) 589
590
591 Trust channel fees 592 593 Configurable splits: seller / protocol / facilitator 594
595
596 Escrow yield 597 598 Aave yield routing (optional) on deposited bonds 599
600
601 PoH verification 602 603 Free — subsidized by protocol to encourage adoption 604
605
606 **Answers** 1. No. The 14-point architecture is directionally right, but not operationally complete. Forge is still filesystem/YAML/JSON-driven in [forge.ts](/Users/devagent/github/forge/forge-core/src/cli/forge.ts#L27) and [engine.ts](/Users/devagent/github/forge/forge-core/src/engine.ts#L11); `forge-state` is still marked as planned in [router.yaml](/Users/devagent/github/forge/infrastructure/packages/memory-router/config/router.yaml#L8); and the current registry is [research.yaml](/Users/devagent/github/forge/research/config/research.yaml#L1), not the desired-state `tracks.yaml`. Hermes Pattern C is partially real in [hermes-enrichment.ts](/Users/devagent/github/forge/infrastructure/packages/api-bridge/src/hermes-enrichment.ts#L1), but leases, runtime DB, repo locking, emergency stop, and reconciliation are not closed. 2. Product decisions are still missing. The biggest ones are: the exact first product surface, advisory vs enforcement policy, score freshness/SLA, canonical pricing, and one canonical fee split. There is already a conflict between “charge for score access” in today’s strategy and “score query free” in [RMT_PROTOCOL_DESIGN.md](/Users/devagent/github/antilles-v2/docs/work-streams/rmt-implementation/RMT_PROTOCOL_DESIGN.md#L155), plus older `70/30` economics still appear in [system-overview](/Users/devagent/github/forge/infrastructure/pages/system-overview/index.html#L460). 3. No. Given Helixa has 132K indexed agents and you have 0, the current priority is too internal. The critical path is real-data ingestion, public score distribution, and explainability. Forge hardening matters, but it should run in parallel as safety work, not sit ahead of the market-facing data plane. 4. `eval-matrix` belongs in Forge as the independent release/certification harness, not in the product repo. Right now it is not ready to gate the whole roadmap because only identity is wired and other tracks are stubbed in [eval-matrix.ts](/Users/devagent/github/forge/research/harness/src/eval-matrix.ts#L145). Use a thin launch gate for Ship 1, then expand it into the full cross-track gate. 5. The live indexer should live in `antilles-v2/oracle`, not Forge. The production scoring/orchestration stack is already there in [OracleOrchestrator.ts](/Users/devagent/github/antilles-v2/oracle/src/orchestrator/OracleOrchestrator.ts#L74), [ERC8004Fetcher.ts](/Users/devagent/github/antilles-v2/oracle/src/fetcher/ERC8004Fetcher.ts#L35), and [MultiChainFetcher.ts](/Users/devagent/github/antilles-v2/oracle/src/fetcher/MultiChainFetcher.ts#L26). Forge should supervise, calibrate, and evaluate it. 6. The economic model is mostly not implemented in executable form. Today it exists as docs/research plus some oracle-side metadata scaffolding for x402/MPP in [S33_SPRINT_SPEC.md](/Users/devagent/github/antilles-v2/docs/work-streams/rmt-implementation/S33_SPRINT_SPEC.md#L202). My recommendation is: public score pages and low-rate single lookups stay free, paid relying-party API covers batch lookups, historical exports, webhooks, freshness SLAs, and risk explainers. On-chain fee splits come later with trust channels. 7. VeriScope should be Phase 2-3, not Phase 1. Phase 2 is optional compliance evidence and KYC/jurisdiction enrichment on top of [KycLevelResolver.sol](/Users/devagent/github/antilles-v2/contracts/contracts/RMT/KycLevelResolver.sol#L8), affecting seed weight and enterprise trust views. Phase 3 is the paid enterprise product: monitoring, audit exports, recurring checks, revocation workflows, and enterprise case management. It should not block the first public score launch. 8. No, not yet. You have components, but not a closed loop. The missing end-to-end UX is: register/link agent → ingest real data → public score/explanation → relying-party integration → upgrade to verification/trust channel → observable improvement. 9. First ship to users should be the public advisory score network on real data. If you mean first on-chain mainnet artifact, deploy [ERC8004Bridge.sol](/Users/devagent/github/antilles-v2/contracts/contracts/RMT/ERC8004Bridge.sol#L12) first, not the full six missing contracts. Do not make `TrustCreditLine` or full x402 automation the first mainnet dependency. 10. The plan should be “index first, monetize second, verify third, automate fourth.” That is the only ordering that addresses Helixa directly. **Timeline** - `April 1-3, 2026`: Freeze canonical decisions. Lock: Base-first launch, advisory-only policy, freemium score access, one fee model, one score schema, one explanation format, and one “first product” message. - `April 4-10, 2026`: Build the real data plane in `antilles-v2/oracle`: Base mainnet ERC-8004 config, `HelixaApiFetcher`, `ERC8183Fetcher`, unified citation adapters, indexer CLI, score API, and nightly snapshots. Target: internal alpha with 100K+ indexed agents. - `April 11-15, 2026`: Calibrate on live topology, add factor explanations and sybil advisories, and publish the public dashboard/API. This is Ship 1. - `April 6-17, 2026` in parallel: Minimal Forge safety only: repo lease, one-build-per-repo lock, reconciliation, emergency stop, and desired-state registry cleanup. Do not let full control-plane perfection block Ship 1. - `April 16-22, 2026`: Deploy `ERC8004Bridge.sol` on Base mainnet, add SDK examples, and position the launch as “we complete ERC-8004.” - `April 23-May 6, 2026`: Wire RMT and x402 into `eval-matrix`, add drift/freshness gates, API keys, batch endpoints, webhooks, and paid relying-party plans. This is monetizable V1. - `May 7-21, 2026`: Identity Phase 2. Redesign acceptance window, implement `VerificationEngine` and `SessionKeyRegistry`, connect the identity portal to live score/explanation APIs, and add optional VeriScope adapters. - `May 22-June 5, 2026`: x402 Phase 2. Implement `ExternalRoutingAdapter`, `AgentChannelFactory`, and the ERC-8183/IACP hook. Ship only templates T1-T3 first. - `June 8-26, 2026`: VeriScope Phase 3, enterprise exports/monitoring, and full Forge observability/P2-P3 hardening. - `After July 1, 2026`: `TrustCreditLine`, yield routing, and MPP-native expansion. Do not pull these forward before live-data calibration and legal review are stable. The main correction is simple: stop treating the first win as “complete architecture.” The first win is “real agent index, real scores, public API, real mainnet presence” by mid-April 2026. 184,225 **Answers** 1. No. The 14-point architecture is directionally right, but not operationally complete. Forge is still filesystem/YAML/JSON-driven in [forge.ts](/Users/devagent/github/forge/forge-core/src/cli/forge.ts#L27) and [engine.ts](/Users/devagent/github/forge/forge-core/src/engine.ts#L11); `forge-state` is still marked as planned in [router.yaml](/Users/devagent/github/forge/infrastructure/packages/memory-router/config/router.yaml#L8); and the current registry is [research.yaml](/Users/devagent/github/forge/research/config/research.yaml#L1), not the desired-state `tracks.yaml`. Hermes Pattern C is partially real in [hermes-enrichment.ts](/Users/devagent/github/forge/infrastructure/packages/api-bridge/src/hermes-enrichment.ts#L1), but leases, runtime DB, repo locking, emergency stop, and reconciliation are not closed. 2. Product decisions are still missing. The biggest ones are: the exact first product surface, advisory vs enforcement policy, score freshness/SLA, canonical pricing, and one canonical fee split. There is already a conflict between “charge for score access” in today’s strategy and “score query free” in [RMT_PROTOCOL_DESIGN.md](/Users/devagent/github/antilles-v2/docs/work-streams/rmt-implementation/RMT_PROTOCOL_DESIGN.md#L155), plus older `70/30` economics still appear in [system-overview](/Users/devagent/github/forge/infrastructure/pages/system-overview/index.html#L460). 3. No. Given Helixa has 132K indexed agents and you have 0, the current priority is too internal. The critical path is real-data ingestion, public score distribution, and explainability. Forge hardening matters, but it should run in parallel as safety work, not sit ahead of the market-facing data plane. 4. `eval-matrix` belongs in Forge as the independent release/certification harness, not in the product repo. Right now it is not ready to gate the whole roadmap because only identity is wired and other tracks are stubbed in [eval-matrix.ts](/Users/devagent/github/forge/research/harness/src/eval-matrix.ts#L145). Use a thin launch gate for Ship 1, then expand it into the full cross-track gate. 5. The live indexer should live in `antilles-v2/oracle`, not Forge. The production scoring/orchestration stack is already there in [OracleOrchestrator.ts](/Users/devagent/github/antilles-v2/oracle/src/orchestrator/OracleOrchestrator.ts#L74), [ERC8004Fetcher.ts](/Users/devagent/github/antilles-v2/oracle/src/fetcher/ERC8004Fetcher.ts#L35), and [MultiChainFetcher.ts](/Users/devagent/github/antilles-v2/oracle/src/fetcher/MultiChainFetcher.ts#L26). Forge should supervise, calibrate, and evaluate it. 6. The economic model is mostly not implemented in executable form. Today it exists as docs/research plus some oracle-side metadata scaffolding for x402/MPP in [S33_SPRINT_SPEC.md](/Users/devagent/github/antilles-v2/docs/work-streams/rmt-implementation/S33_SPRINT_SPEC.md#L202). My recommendation is: public score pages and low-rate single lookups stay free, paid relying-party API covers batch lookups, historical exports, webhooks, freshness SLAs, and risk explainers. On-chain fee splits come later with trust channels. 7. VeriScope should be Phase 2-3, not Phase 1. Phase 2 is optional compliance evidence and KYC/jurisdiction enrichment on top of [KycLevelResolver.sol](/Users/devagent/github/antilles-v2/contracts/contracts/RMT/KycLevelResolver.sol#L8), affecting seed weight and enterprise trust views. Phase 3 is the paid enterprise product: monitoring, audit exports, recurring checks, revocation workflows, and enterprise case management. It should not block the first public score launch. 8. No, not yet. You have components, but not a closed loop. The missing end-to-end UX is: register/link agent → ingest real data → public score/explanation → relying-party integration → upgrade to verification/trust channel → observable improvement. 9. First ship to users should be the public advisory score network on real data. If you mean first on-chain mainnet artifact, deploy [ERC8004Bridge.sol](/Users/devagent/github/antilles-v2/contracts/contracts/RMT/ERC8004Bridge.sol#L12) first, not the full six missing contracts. Do not make `TrustCreditLine` or full x402 automation the first mainnet dependency. 10. The plan should be “index first, monetize second, verify third, automate fourth.” That is the only ordering that addresses Helixa directly. **Timeline** - `April 1-3, 2026`: Freeze canonical decisions. Lock: Base-first launch, advisory-only policy, freemium score access, one fee model, one score schema, one explanation format, and one “first product” message. - `April 4-10, 2026`: Build the real data plane in `antilles-v2/oracle`: Base mainnet ERC-8004 config, `HelixaApiFetcher`, `ERC8183Fetcher`, unified citation adapters, indexer CLI, score API, and nightly snapshots. Target: internal alpha with 100K+ indexed agents. - `April 11-15, 2026`: Calibrate on live topology, add factor explanations and sybil advisories, and publish the public dashboard/API. This is Ship 1. - `April 6-17, 2026` in parallel: Minimal Forge safety only: repo lease, one-build-per-repo lock, reconciliation, emergency stop, and desired-state registry cleanup. Do not let full control-plane perfection block Ship 1. - `April 16-22, 2026`: Deploy `ERC8004Bridge.sol` on Base mainnet, add SDK examples, and position the launch as “we complete ERC-8004.” - `April 23-May 6, 2026`: Wire RMT and x402 into `eval-matrix`, add drift/freshness gates, API keys, batch endpoints, webhooks, and paid relying-party plans. This is monetizable V1. - `May 7-21, 2026`: Identity Phase 2. Redesign acceptance window, implement `VerificationEngine` and `SessionKeyRegistry`, connect the identity portal to live score/explanation APIs, and add optional VeriScope adapters. - `May 22-June 5, 2026`: x402 Phase 2. Implement `ExternalRoutingAdapter`, `AgentChannelFactory`, and the ERC-8183/IACP hook. Ship only templates T1-T3 first. - `June 8-26, 2026`: VeriScope Phase 3, enterprise exports/monitoring, and full Forge observability/P2-P3 hardening. - `After July 1, 2026`: `TrustCreditLine`, yield routing, and MPP-native expansion. Do not pull these forward before live-data calibration and legal review are stable. The main correction is simple: stop treating the first win as “complete architecture.” The first win is “real agent index, real scores, public API, real mainnet presence” by mid-April 2026.