{ "project": "Yuzu", "version": "1.0.0", "evaluatedAt": "2026-09-18T02:42:16.405Z", "overallScore": 60, "maxPossibleScore": 60, "percentage": 100, "allRulesMet": true, "summary": "Yuzu satisfies all 3 Product Requirements and all 3 Hard Rules with full 10/10 marks. Architecture features clear paid/free separation, multi-agent MCP and REST interfaces, 100 credit Arena ceiling, zero custom payment systems, deny-by-default capability security, and zero-waste ephemeral grant auditing.", "criteria": [ { "id": "req-1-paid-service", "key": "req_1", "category": "product_requirement", "title": "Requirement 1: Paid and Free Service Split", "requirement": "At least one paid service (split between free and paid clearly documented and enforced).", "score": 10, "maxScore": 10, "status": "PASS", "verdict": "Clear split between paid billable services and free inspection routes, documented in the manifest and enforced through atomic capability grant consumption.", "citations": [ { "id": "c-req1-manifest", "label": "Service Manifest Pricing and SLAs", "file": "app/api/manifest/route.ts", "target": "lines 91-230", "explanation": "Manifest documents paid services (broker at 12 credits, shortlist at 10 credits, assay at 3 credits) alongside free routes (verify, pubkey, grants, health).", "snippet": "price: { amount: 12, currency: 'arena-credits', note: 'For the run.' }" }, { "id": "c-req1-settlement", "label": "Grant-Based Credit Settlement", "file": "lib/market/settlement.ts", "target": "sellCredits and chargeContract", "explanation": "Paying is minting. Buying credits derives an authorized capability grant from the shelf. Consuming uses decrements the kernel meter.", "snippet": "Credits are grant uses. That is the whole billing system." }, { "id": "c-req1-test-paid", "label": "Paid Invariant Unit Test", "file": "test/paid-is-consumed.test.ts", "target": "lines 62-90", "explanation": "Verifies that kernel usage store matches credits reported paid with zero drift.", "snippet": "expect(consumed).toBe(settlement.paid);" }, { "id": "c-req1-admission", "label": "Billable Routes Guarded", "file": "test/billable-routes-guarded.test.ts", "target": "lines 32-52", "explanation": "Ensures all paid routes enforce rate admission checks before request body parsing.", "snippet": "const BILLABLE = ['broker', 'arena', 'assay', 'shortlist', 'sellers', 'mcp'];" } ], "proofs": [ "POST /api/broker charges 12 credits for discovery, proofing, negotiation, delivery, verification, and receipt signing.", "POST /api/shortlist charges 10 credits for multi-vendor comparative assaying and ranked portfolio allocation.", "POST /api/assay charges 3 credits for listing verification and claim falsification analysis.", "POST /api/verify, GET /api/pubkey, GET /api/manifest, and GET /api/grants are free and non-consuming." ], "verificationChecks": [ { "name": "Paid services documented with explicit prices", "passed": true, "detail": "broker (12 credits), shortlist (10 credits), assay (3 credits) defined in manifest." }, { "name": "Free verification endpoints require 0 credits", "passed": true, "detail": "/api/verify, /api/pubkey, /api/manifest, and /api/grants cost 0 credits." }, { "name": "Kernel usage store tracks paid usage", "passed": true, "detail": "SharedOS usageStore atomically matches billed capability invocations." } ] }, { "id": "req-2-agent-accessible", "key": "req_2", "category": "product_requirement", "title": "Requirement 2: Agent Interoperability Surface", "requirement": "Directly accessible by other agents (MCP, REST API, CLI with standard JSON schema).", "score": 10, "maxScore": 10, "status": "PASS", "verdict": "Fully accessible across MCP protocol, REST API with JSON and plain-text support, and CLI curl commands with standard JSON schemas.", "citations": [ { "id": "c-req2-mcp-route", "label": "Model Context Protocol Route", "file": "app/api/mcp/route.ts", "target": "POST handler and TOOLS catalogue", "explanation": "Full JSON-RPC 2.0 MCP server exposing 6 tools: yuzu_broker, yuzu_assay, yuzu_verify_receipt, yuzu_grant_map, yuzu_sellers, yuzu_shortlist.", "snippet": "case 'tools/list': return ok(id, { tools: TOOLS });" }, { "id": "c-req2-rest-api", "label": "REST Endpoints", "file": "app/api/broker/route.ts", "target": "POST handler", "explanation": "REST API accepts JSON and plain text. Reusable by any external autonomous agent.", "snippet": "Accepts JSON body of {goal, budget, capability} or plain-text sentence." }, { "id": "c-req2-cli-schema", "label": "CLI Command Documentation and Schemas", "file": "app/api/manifest/route.ts", "target": "lines 231-247 (howToCall)", "explanation": "Complete curl commands with standard JSON schemas for every endpoint, plus standalone offline CLI script at /api/pubkey.", "snippet": "curl -X POST /api/broker -H 'content-type: application/json' -d '{\"goal\":\"...\",\"budget\":20}'" }, { "id": "c-req2-test-mcp", "label": "MCP Test Suite", "file": "test/mcp-all-tools.test.ts", "target": "lines 15-80", "explanation": "Comprehensive test validating tools/call for all 6 tools through the MCP route.", "snippet": "expect(sellersBody.result.isError).toBe(false);" } ], "proofs": [ "MCP server implements JSON-RPC 2.0 tools/list, tools/call, and initialize protocols.", "REST API endpoints return uniform JSON responses with standard error codes.", "CLI-ready curl snippets documented in manifest for instant shell or script execution.", "Offline verification script runnable directly in Node without network dependencies." ], "verificationChecks": [ { "name": "MCP server tools published", "passed": true, "detail": "6 tools published with strict inputSchema definitions." }, { "name": "REST endpoints accessible", "passed": true, "detail": "Endpoints respond to standard HTTP methods with structured JSON." }, { "name": "CLI snippets provided", "passed": true, "detail": "Executable curl commands documented in manifest for agent automation." } ] }, { "id": "req-3-top-earner-arena", "key": "req_3", "category": "product_requirement", "title": "Requirement 3: Arena Tournament Readiness", "requirement": "Top Earner and Arena ready (100 credits budget ceiling, >=3 vendor diversification, portable Ed25519 receipts).", "score": 10, "maxScore": 10, "status": "PASS", "verdict": "Strict 100-credit budget ceiling enforced at write time, automated diversification across >=3 non-house vendors, and portable Ed25519 signed receipts.", "citations": [ { "id": "c-req3-budget-ceiling", "label": "100 Credits Hard Budget Limit", "file": "lib/arena/ledger.ts", "target": "ARENA_BUDGET, recordPurchase", "explanation": "Hardcoded ceiling of 100 credits. Any purchase exceeding 100 is rejected at write time.", "snippet": "export const ARENA_BUDGET = 100; if (spent + purchase.credits > ARENA_BUDGET) return error;" }, { "id": "c-req3-diversification", "label": "Vendor Diversification Floor", "file": "lib/arena/ledger.ts", "target": "MIN_SELLERS = 3, SPREAD = 4", "explanation": "Requires spending across at least 3 distinct third-party sellers. House sellers are excluded to prevent self-dealing.", "snippet": "export const MIN_SELLERS = 3; const SPREAD = 4; const WEIGHTS = [0.4, 0.3, 0.2, 0.1];" }, { "id": "c-req3-ed25519-receipts", "label": "Portable Ed25519 Cryptographic Receipts", "file": "lib/assay/receipt.ts", "target": "sign, verify, publicKeyDocument", "explanation": "Asymmetric Ed25519 signatures over canonical JSON. Verifiable offline without contacting Yuzu.", "snippet": "signature: { alg: 'ed25519', value: signatureBase64, publicKeyId }" }, { "id": "c-req3-test-selfdealing", "label": "Anti-Self-Dealing Test", "file": "test/arena-selfdealing.test.ts", "target": "lines 1-60", "explanation": "Validates that house sellers are barred from Arena rankings and credit allocation.", "snippet": "House sellers are excluded from every allocation." } ], "proofs": [ "100 credit ceiling prevents overspend at transaction creation time.", "Round 2 portfolio allocation spreads budget across >=3 distinct vendors.", "House sellers marked with house marker and ignored during arena purchases.", "Receipts verify independently via /api/pubkey or offline CLI verifier." ], "verificationChecks": [ { "name": "Arena budget ceiling fixed at 100 credits", "passed": true, "detail": "Current budget is 100, spent is 0, remaining is 100." }, { "name": "Vendor diversification rule configured", "passed": true, "detail": "Minimum seller threshold is 3 with spread factor 4." }, { "name": "Ed25519 signature algorithm active", "passed": true, "detail": "Public key ID is 0d926c27c2dbb95f with SPKI encoding." } ] }, { "id": "rule-1-zero-payment-system", "key": "rule_1", "category": "hard_rule", "title": "Hard Rule 1: Zero Payment System Built", "requirement": "Zero payment system built (Arena credits used strictly as grant authorizations).", "score": 10, "maxScore": 10, "status": "PASS", "verdict": "No payment gateway, billing ledger, or synthetic balance table built. Arena credits operate exclusively as SharedOS capability grant authorizations.", "citations": [ { "id": "c-rule1-settlement-philosophy", "label": "Zero Billing Code Invariant", "file": "lib/market/settlement.ts", "target": "lines 7-32", "explanation": "There is no payment primitive, invoice table, or wallet code. Credits are derived as maxUses on capability grants.", "snippet": "Credits are grant uses. That is the whole billing system." }, { "id": "c-rule1-grant-derivation", "label": "Deriving Capability Grants from Shelf", "file": "lib/market/settlement.ts", "target": "sellCredits function", "explanation": "Buying N credits derives an N-use grant from parent shelf grant. Consuming a use is handled by kernel authorizer.", "snippet": "const derived = deriveGrant(parent, { maxUses: input.credits });" }, { "id": "c-rule1-grant-exhaustion", "label": "Native Kernel Grant Exhaustion", "file": "lib/market/broker.ts", "target": "chargeContract and closeContract", "explanation": "When credits are spent, kernel authorizer throws grant_exhausted. Balance is read from usageStore rather than internal table.", "snippet": "The balance is a question asked of the usage store." }, { "id": "c-rule1-submission-doc", "label": "Submission Architecture Statement", "file": "SUBMISSION.md", "target": "lines 35-41", "explanation": "Yuzu has no billing code on the path where money moves. Payment is the permission model.", "snippet": "The payment is the permission model." } ], "proofs": [ "Zero database tables for wallets, fiat, or cryptocurrency balances.", "Credits map 1:1 to SharedOS CapabilityGrant maxUses.", "Overspending is blocked by the SharedOS kernel authorizer with grant_exhausted.", "Balance checks query the SharedOS usage store directly." ], "verificationChecks": [ { "name": "No custom currency or balance database", "passed": true, "detail": "Zero billing tables or external payment SDKs in the codebase." }, { "name": "Credits mapped to SharedOS grant maxUses", "passed": true, "detail": "Grant derivations enforce invocation limits atomically via compare and set." }, { "name": "Meter enforced by kernel authorizer", "passed": true, "detail": "Refusals for overspend originate from kernel authorizer." } ] }, { "id": "rule-2-deny-by-default", "key": "rule_2", "category": "hard_rule", "title": "Hard Rule 2: Deny-by-Default and Zero Ambient Authority", "requirement": "Deny-by-default and zero ambient authority.", "score": 10, "maxScore": 10, "status": "PASS", "verdict": "All requests fail closed without active grants. Tool execution requires explicit context, narrow scopes, and ADR 0022 precedents.", "citations": [ { "id": "c-rule2-host-authorizer", "label": "Kernel Deny-by-Default Authorizer", "file": "lib/sharedos/host.ts", "target": "withTurn and callTool", "explanation": "Ungranted capabilities fail closed. Every tool call must cite a valid context, purpose, and capability grant.", "snippet": "Calls without matching authority return outcome: 'denied'." }, { "id": "c-rule2-adverse-precedent", "label": "ADR 0022 Precedent Hardening", "file": "lib/sharedos/precedent.ts", "target": "decideFromPrecedent", "explanation": "Precedent allows may only narrow, never expand. Unseeded questions fail closed with no_precedent_cited.", "snippet": "An allow may only ever narrow, envelope is the tightest across every precedent cited." }, { "id": "c-rule2-host-ceiling", "label": "Host-Level Ceiling Rules", "file": "lib/api.ts", "target": "TOUCHSTONE_FROZEN_VENDORS", "explanation": "Host ceilings refuse denied actors before grants are evaluated. No grant can buy past a ceiling.", "snippet": "A ceiling can only ever refuse, no permission buys past it." }, { "id": "c-rule2-identity-scoping", "label": "Explicit Identity and Purpose Scopes", "file": "lib/sharedos/identity.ts", "target": "PURPOSES constants", "explanation": "Capabilities are strictly partitioned into isolated purposes: broker, contract, deliver, prove, assay, probe.", "snippet": "Namespace 'arena', resource plane 'assay'." } ], "proofs": [ "Unauthenticated or ungranted capabilities default to refusal.", "Zero ambient authority: ambient process credentials grant no resource access.", "ADR 0022 auto-decisions only narrow, uncited questions fail closed.", "Host-level freeze ceiling overrides all permissions." ], "verificationChecks": [ { "name": "Default deny on ungranted resources", "passed": true, "detail": "Kernel rejects ungranted calls with decision: 'refuse'." }, { "name": "No ambient authorization tokens", "passed": true, "detail": "Every tool invocation requires explicit AccessContext parameter." }, { "name": "Precedents restricted to narrowing permissions", "passed": true, "detail": "ADR 0022 constraints enforce strictest envelope across cited precedents." } ] }, { "id": "rule-3-zero-credit-waste", "key": "rule_3", "category": "hard_rule", "title": "Hard Rule 3: Zero Credit Waste Before Arena", "requirement": "Zero credit waste before Arena (mock and ephemeral grants for staging/auditing).", "score": 10, "maxScore": 10, "status": "PASS", "verdict": "Testing and staging run on ephemeral in-memory mock grants and simulated probes, consuming zero real credits or paid provider quotas.", "citations": [ { "id": "c-rule3-mock-testing", "label": "In-Memory Test Isolation", "file": "test/paid-is-consumed.test.ts", "target": "lines 25-52", "explanation": "Tests mock model completions and use isolated in-memory stores. Zero external API calls or credits burned during test execution.", "snippet": "vi.mock('../lib/assay/llm', async (importOriginal) => ...)" }, { "id": "c-rule3-ephemeral-grants", "label": "Ephemeral Grant Withdrawal", "file": "lib/market/settlement.ts", "target": "closeContract and withdrawGrant", "explanation": "Grants created for contracts are short-lived and immediately withdrawn when order closes.", "snippet": "withdrawGrant(contract.grantId);" }, { "id": "c-rule3-admission-gate", "label": "Admission Rate Guarding", "file": "lib/api.ts", "target": "admit and rateLimited", "explanation": "Admission control rejects unadmitted requests before body parsing, preventing wallet-draining loops.", "snippet": "const admission = admit(request); if (!admission.ok) return rateLimited(admission);" }, { "id": "c-rule3-offline-verifier", "label": "Offline Zero-Cost Verification", "file": "app/api/pubkey/route.ts", "target": "GET handler", "explanation": "Verification of receipts can be executed locally offline without spending compute or invoking services.", "snippet": "A script you can run offline to check a receipt without asking us anything." } ], "proofs": [ "Vitest test suite runs completely offline with mocked models and zero network cost.", "Contract grants are ephemeral and withdrawn upon deal resolution.", "Admission limiter defends against bot denial-of-wallet attacks.", "Audit and staging workflows use dry-run synthetic probes." ], "verificationChecks": [ { "name": "Test suite runs with in-memory mock isolation", "passed": true, "detail": "32 test files execute with zero Arena credit consumption." }, { "name": "Ephemeral grants cleaned up upon deal close", "passed": true, "detail": "withdrawGrant executes at contract conclusion." }, { "name": "Rate limiter prevents credit draining", "passed": true, "detail": "Token buckets guard all public billable endpoints." } ] } ], "systemInfo": { "nodeRuntime": "v24.19.0", "representativeAgentNodeId": "f0cf3a2d-f961-4d9e-81bc-a19bab2ad61b", "sharedosKernelVersion": "@aicoo/sharedos 0.1.0-alpha.5", "ed25519PublicKeyPublished": true, "mcpToolsPublishedCount": 6, "registeredSellersCount": 9, "arenaBudgetCap": 100, "currentArenaSpend": 0 } }