GOLDTRINITY — War Room & Full Agent Orchestration Studio v2.0

Factor MAD Pair

Agent-Builder

Risk Officer (CRO)

Macro-Researcher

Sentiment-Analyst

Resolution-Differ

News-Analyst

Bull-Researcher

Alpha-Miner

VPIN-Analyst

Conductor (CIO)

Reserved slot (4)

Process: Blindspot Loop

EXECUTION-ANALYST

Collector PM Crypto

Data-Steward

Asymmetric freedom /
sealed evaluator (AQuA)

Reserved slot (4)

CLAUDE.md routing
+ guardrail_gateway

▶ AGENT HUD

AERON

CHARTERED

Conductor (Fund Manager / CIO)

COMMAND · fable-5 · UPPER TIER

AI Gateway Route:

gateway.ai.cloudflare.com/v1/.../anthropic

Primary KPI: routing / daily brief

Live Animation: SITTING (TYPING)

AGENT STUDIO & SETTINGS SUITE · Fleet Config · LLM Models & Secret Keys · Modular Avatar Genome

Agent ID / Codename

Agent Display Name

Organizational Role

Desk / Department

Floor Tier

Lifecycle Status

Primary Performance KPI

LLM Foundation Model

🔐 SECRET KEYS & INFERENCE GATEWAY (SECURE CLIENT VAULT)

API Key / Token (Masked)

Custom Base URL (Optional)

Account ID / Org ID (Optional)

Secret Key (Optional)

Private Key (Optional - RSA-PSS / ECDSA Signatures for Kalshi/Venue Signing)

LIVE 96x96 PIXEL GENOME

MODULAR COMPOSITOR

Gender / Archetype

Skin Tone

Head & Jaw Shape

Eyes & Visor

Hair Style

Hair Color

Shirt / Upper Body

Pants / Lower Body

Shoes / Footwear

Accessory 1 (Eyewear/Hat)

Accessory 2 (Watch/Socks)

🖼️ MANUAL SPRITE / AVATAR FILE UPLOAD (PNG/SVG)

AGENT DASHBOARD · roster & orchestration KPIs · fed by /api/pulse LIVE TELEMETRY AGGREGATOR

Org structure (a real fund, translated to agents)

PRINCIPAL

(human, T3)

CONDUCTOR

(fund manager / CIO)

RESEARCH

(analysts)

QUANT

(sims)

RISK

(veto)

EXECUTION

(daemons)

OPERATIONS

(builder)

Org structure (start 20 active, scale via agent-builder)

Capability Gaps (GAP-SENTINEL)

Coverage Gap: IBKR Greeks
Failure Pattern: NegRisk misread
Failure Pattern: NegRisk · sports
Failure Pattern: NegRisk · politics

Gap:

backlog review

How new agents get born (blindspot loop)

BLINDSPOT

detection

→

CONDUCTOR

authorization

→

AGENT-BUILDER

spawn in shadow

7 days shadow · PRM-scored · beats eval → promoted · fails → archived

The roster (start 20 active, scale via agent-builder)

Command:

● conductor (CIO)
● agent-builder (Head talent)

Research Desk:

● macro-researcher
● news-analyst
● sentiment-analyst
● resolution-differ
● bull-researcher
● bear-researcher
● RESEARCH-LIBRARIAN

Quant Desk:

● alpha-miner
● factor-advocate
● factor-skeptic
● quant-simulator
● forecaster
● vpin-analyst
● REGIME-DETECTOR

Risk & Compliance:

● risk-officer
● prm-auditor
● compliance-agent
● allocator

● executor (Head trader · none)
● collector_pm/crypto (Fill forensics)
● graph-engine

Operations & Capability:

● ops-reporter · intel-watcher
● GAP-SENTINEL (Capability Intel)
● HR-REVIEWER (Performance)
● DATA-STEWARD · EXECUTION-ANALYST

✦ Highlighted CYAN items indicate v2 Addendum system additions.

Capability Gaps (GAP-SENTINEL)

– Coverage Gap: IBKR Greeks
– Failure Pattern: NegRisk misread
– Failure Pattern: NegRisk misread
– Failure Pattern: Excess risk
– Failure Pattern: NegRisk misread

Portfolio Allocation

Family

Target

Weights

politics

40%

→ 25%

sports

30%

25%

crypto

50% ▲

40%

macro (prop.)

10%

Proposed

● Target weight
● Adjustment +3%
● Proposed adjustment: −33%

Skill libraries promotion rules

builtin→trial→accepted→rejected

auto-promote |ρ| > 0.3 · demote < 0.05 · one owner per library

Quant Simulation

CPCV

Test

Monte Carlo

Test

Walk-Forward

Test

Sealed evaluator ARMED

Model→compute map

Claude Max (6 seats)
Gemini Ultra
Kimi K3
Alibaba
HuggingFace

Agent Evals (HR-REVIEWER)

3 fails → PIP

+1 fail → Offboarding

PIP tighter scope

pass → Not Offboarding

weekly · quality × veto rate × PRM calibration