Genesis Conductor × UCP — Integration Spec & Model Attainment Decision Matrix

<aside> 🧭

Single-path output. Four primitives — EulerCycleAttestor.sol — Clock Spring Redesign (supersedes EulersIdentitySynthesis.sol), EulerCycleAttestor.sol — Full Codebase & Test Suite, ‣, and the Self-Evolving Agent Retraining Worker — composed into a UCP-shaped capability/extension fabric, routed by Ambient Agent Access Layer — Product Spec, Schema, and Skeleton with Greg as architect-tier discovery vertex. The decision matrix below selects the model for any task by task_type × required_confidence × policy_tier × budget.

</aside>

<aside> 🔗

Canonical cluster anchors (previewable). Yennefer cluster · TAEX · Retraining loop · §6.17 · Integrations registry · Integrative Flow.

‣ · ‣ · ‣ · ‣ · ‣ · TAEX Intent Router — Agent-Based Routing with Dissonance Memory MCP · ‣ · ‣ · ‣

</aside>

Layer 1 — Solution

1. Topology

flowchart TB
	User["User / Caller"] -->|request| AAL["Ambient Access Layer<br/>request_type → primaryAgent + extensions[]"]
	AAL -->|architecture / spec_review| Greg["Greg<br/>(architect vertex)"]
	AAL -->|maru_reframe<br/>(no_win=true)| Maru["/maru<br/>(reframe extension)"]
	AAL -->|prompt_eval / prompt_optimize| Retrain["Self-Evolving<br/>Retraining Worker"]
	AAL -->|implementation| Workers["Worker fleet<br/>(Kiro, Codex, Claude Code,<br/>Gemini CLI, Chromium Gemini VM)"]
	Greg -->|invariants ratified| Workers
	Maru -->|reframe| Workers
	Workers -->|output + reportHash| Retrain
	Retrain -->|pass| ECA["EulerCycleAttestor v2<br/>(escapement + KVDF NFT mint)"]
	Retrain -->|fail| Cancel["cancelStaleCycle"]
	ECA -->|settleCycle → mint NFT| Capsule["Soul Capsule<br/>tokenId = canonical prompt-version handle"]
	Capsule -->|reference| AAL
	Capsule -->|reference| WOPS["W-OPS<br/>workspace state"]
	Capsule -->|reference| Tracker["🚀 Deployment Tracker"]

2. UCP capability / extension mapping

UCP defines: capabilities (discrete primitives), extensions (capability augmentations), discovery (agents query profiles), and transport-agnostic binding (REST / MCP / A2A). Genesis Conductor inherits the shape; verbs change from commerce to model-attainment.

UCP concept Genesis Conductor equivalent Surface
Capability: product_discovery Capability: model_attainment AAL POST /v1/tasks
Extension: discount Extension: no_win_reframe /maru fires when Guardian R > 0.4
Extension: tax_calc Extension: gas_bound_quote armTick(gasBoundWei)
Discovery (agent queries) Discovery (AAL router) request_type → primaryAgent + extensions[]
Transport: REST / MCP / A2A Transport: AAL / MCP / A2N Same matrix, transport-agnostic

3. Decision matrix for model attainment

Given a task, the matrix selects: primary agent, suggested model class, UCP extensions that fire, policy tier, and cycle budget. The Architect vertex (Greg or /maru-Architect) emits ≥3 candidate routings; the matrix is the prior.

Task type Primary agent Model class Required confidence Policy tier UCP extensions Cycle / settlement Architecture authorship / spec review Greg Claude Sonnet 4.5 verified standard agent_charter_lookup, gas_bound_quote 1 cycle, NFT mint on ratification
Constraint reframe (no-win detected) Pareto + /maru Claude Sonnet 4.5 plausible standard no_win_reframe 1 cycle, no settlement (advisory) Reframe with commit hooks (mode=execute) Pareto + /maru → Greg adjudication Claude Sonnet 4.5 verified prod_sensitive no_win_reframe, agent_charter_lookup 1–3 cycles, NFT mint on commit
Feature implementation Kiro Claude Sonnet 4.5 plausible standard gas_bound_quote 2 cycles, NFT mint on PR-merge Terminal / ops automation Codex GPT-5 plausible standard terminal_ops 1 cycle, NFT mint on green CI
Browser automation Chromium Gemini VM (W-13) Gemini 2.5 Pro plausible standard browser_runtime 2 cycles, NFT mint on artifact capture Prompt eval / regression Self-Evolving Retraining Worker multi-grader ensemble verified standard prompt_eval 3 cycles (eval pass ratio gate)
Prompt promotion to ledger Retraining Worker → Greg adjudication multi-grader + architect override verified prod_sensitive prompt_eval, prompt_promote, agent_charter_lookup 3–5 cycles, NFT mint on Greg ratify Cross-domain synthesis gc-synthesis (W-14) GPT-5 / Gemini 2.5 Pro plausible standard synthesis 2 cycles, NFT mint on graph-write
Public agentic invocation (.sol-bound) Pareto via maru.sol Claude Sonnet 4.5 plausible standard (free tier) no_win_reframe, public_invoke 1 cycle, no settlement (free tier) On-chain settlement only EulerCycleAttestor v2 n/a (contract) verified prod_sensitive cycle_settlement 1 cycle, KVDF NFT mint mandatory

Selection rule (deterministic):

  1. Match task_type row.
  2. If policy_tier=prod_sensitive is required by caller or by row default, escalate cycle budget to 5 and require Greg adjudication.
  3. If Guardian risk score R > 0.4 on any candidate, auto-fire the no_win_reframe extension (i.e. delegate to /maru).
  4. If required_confidence=verified, the row's settlement step is mandatory — no NFT mint, no attainment.
  5. Auditor drift > 0.15 on emission → re-enter Cycle 1.

4. EulerCycleAttestor as settlement primitive

The v1 redesign explicitly retires the EulersIdentitySynthesis reasoning-engine-as-contract pattern. On-chain ONLY the escapement lives — a fail-closed cadence kernel with a five-step lifecycle: