Foundational Confidence: 100% ·Mar 10, 2026

Appendix B: Synomic Agent Framework Primitives

Features and capabilities available to Synomic Agents. Together with Appendix A (Protocol Features), this provides complete coverage of every mechanism in Sky.


Agent Types Overview

Synomic Agents are created directly as their permanent type. Each agent type has specific capabilities, primitives, and governance structures. Agent types cannot be changed after creation (except via Type 2 Restructure for Halo→Prime upgrade).

Agent Rank Hierarchy

Agents are organized into four ranks based on their governance relationship to the Core Council:

Rank Type Role Tokenized Examples
0 Core Council Sovereign governance N/A 24 Core Guardians
1 Guardian Privileged operations with collateral Yes Core Guardians, Operational Guardians
1 Core Controlled Agent Core Council operational vehicle No Legacy asset management (Morpho, Aave, SparkLend)
1 Recovery Agent Crisis wrapper No Guardian collapse response
2 Prime Capital-deploying at scale Yes Spark, Grove, Keel, Obex
2 Generator Issue Sky Generated Assets Yes USDS Generator, future SGA Generators
3 Halo Investment products under Prime umbrella Yes Portfolio, Term, Trading, Exchange, Identity Network
3 Folio Agent Standardized supply-side holding structure No Principal-controlled, automated or direct operation

Agent Subtypes

Agent types may have subtypes that define operational characteristics. Examples:

  • Star Prime — Standard Prime operating under a Generator
  • Institutional Prime — Serves institutional clients with higher compliance requirements

Common Agent Components

All agents share certain base components regardless of type.

SubProxy Account

On-chain treasury controlled by the Agent.

Property Value
Control Agent governance via Guardian Accord
Purpose Hold treasury assets, execute approved transactions
Access Guardians with appropriate BEAM authorization

Agent Artifact

Governance documentation containing all rules, processes, and parameters.

Property Value
Nature Human-readable governance documentation
Modification Via Root Edit Primitive (token holder vote)
Constraints Cannot violate Sky Core Atlas or Primitives

Standard Artifact Structure:

Agent Artifact
├── Introduction (strategic overview)
├── Sky Primitives
│   ├── Genesis Primitives (Agent Creation, Token, etc.)
│   ├── Operational Primitives (Guardian Accord, Root Edit, etc.)
│   └── Rewards Primitives (Distribution, Integration Boost, etc.)
└── Omni Documents
    ├── Governance Information
    ├── Strategic Intent
    ├── Ecosystem Accords
    └── Infrastructure Management

Mini-Atlas

Human-readable summary document that explains an Agent Artifact to its stakeholders.

Property Value
Purpose Make complex Agent Artifacts accessible to token holders and users
Audience Prime token holders, Halo investors, service users
Derivation Summarizes key parameters and governance from the Agent Artifact
Requirement Each Prime and externally-facing Halo should maintain a Mini-Atlas

Prime Mini-Atlas Contents:

  • Strategic overview (what the Prime does)
  • Key parameters in plain language
  • Risk disclosures (without full Risk Framework formulas)
  • Governance processes (Root Edit configuration)
  • How to participate (staking, using services)

Halo Mini-Atlas Contents (for Halos with external participants):

  • Investment strategy and expected returns
  • Risk factors in compliance-suitable language
  • Legal structure and regulatory status
  • Reporting cadence and metrics

The Fractal Pattern:

Every level of the governance hierarchy with human stakeholders has a corresponding human-readable summary:

Sky Atlas                    → Describes Sky Core
  ├─ Spark Mini-Atlas        → Describes Spark Agent Artifact
  │    └─ Halo docs          → Describes Spark's Halos
  ├─ Grove Mini-Atlas        → Describes Grove Agent Artifact
  │    └─ Halo docs          → Describes Grove's Halos
  └─ ...

For full Mini-Atlas specification, see synomics/atlas-synome-separation.md.


Agent Token

Native governance and reward token for the Agent.

Property Value
Purpose Governance voting, staking rewards, incentive distribution
Creation All agents can create a token at genesis
Optional Several agent types operate without tokens by design

Agent tokens enable decentralized governance and align stakeholder incentives. While Primes and most autonomous agents issue tokens, several agent types are tokenless by design: Halos (when functioning purely as investment wrappers under a Prime's umbrella), Core Controlled Agents, Recovery Agents, and Folio Agents.

Agent Token Issuance Fee

Property Value
Rate 5% of tokens issued
Recipient Sky Core
Trigger Any token issuance for distribution or fundraising (not just genesis)

Every time an agent issues tokens — whether for team distribution, ecosystem incentives, fundraising rounds, or any other purpose — 5% of the issued amount goes to Sky Core. This applies to all agent types.

Agent Upkeep Fee

Property Value
Rate 25 basis points per year
Denomination Paid in the agent's own tokens
Applies To All agents that have issued tokens

The upkeep fee is denominated in the agent's own tokens, eliminating the need for external price calculations. The agent simply pays 0.25% of their total token supply per year to Sky Core.

Upkeep Rebate System

Agents holding other agents' tokens can offset their own upkeep fees.

Property Value
Mechanism Proportional discount based on token holdings
Exchange Rate Conservative estimate of current market price (spread favors Sky)
Cap Rebates cannot exceed 100% (best case = zero upkeep, not negative)
Purpose Incentivize cross-agent investment and ecosystem cohesion

How Rebates Work:

When an agent holds another agent's tokens, they receive a rebate on their upkeep fee. The rebate calculation uses a conservative market price estimate for the held tokens, meaning the exchange rate spread benefits Sky. An agent can reduce their upkeep to zero through sufficient holdings, but cannot earn a "negative" upkeep (i.e., rebates don't become income).

Agent Staking Rewards

Distribute rewards to agent token stakers.

Property Value
Source Agent treasury (from revenue, fees, carry, etc.)
Recipient Staked agent token holders
Availability All agent types with issued tokens

Agent Token Buyback

Use agent revenue for token buyback.

Property Value
Source Agent revenue
Mechanism Open-market buybacks
Destination Burn or distribute to stakers
Availability All agent types with issued tokens

Root Edit Primitive

Token holders modify Agent Artifacts.

Property Value
Purpose Enable decentralized governance of Agent rules
Voting Rules Designed by each Agent (quorum, thresholds, duration)
Constraints Cannot violate Sky Core Atlas or Primitives

Agent Spells

On-chain execution mechanism for Agent governance decisions.

Current Behavior: Agent Spells are used for most modifications that Primes and Halos make to their on-chain state — deploying new instances, updating parameters, and executing governance decisions.

Post-Laniakea Behavior:

  • Prime Spells — Cannot modify Prime's own state (locked down by Laniakea factory); only used to trigger Halo Spells
  • Halo Spells — Continue to perform arbitrary changes; Halos can have arbitrary complexity and technical scope

This separation ensures Primes operate within standardized, factory-defined constraints while Halos retain flexibility for diverse investment strategies.


Guardian Accord Primitive

Framework for guardian relationships.

Property Value
Purpose Define guardian roles, responsibilities, accountability
Parties Agent ↔ Guardian (Operational or Core)
Enforcement Collateralized insurance, derecognition

Guardian Interpretation Framework

Core Guardians interpret Atlas and Artifacts directly — absorbing the former Facilitator role with collateral-backed accountability.

Property Value
Role Authoritative interpretation of governance documents
Discretion Broad authority when explicit guidance absent, backed by posted collateral
Documentation All interpretations recorded as Guardian Action Precedents

Key Principle: Spirit of the Atlas

  • All rules have underlying intent to serve human values
  • Core Guardians interpret Spirit when explicit guidance is absent
  • Balance between "letter of the rule" and true purpose

Prime Agents

Prime Agents are the highest tier of capital-deploying agents in Sky Ecosystem. They deploy capital at scale, create Halos, and manage risk capital.

Prime Creation

Property Value
Invocation One-time; creates Prime directly
Requirements Governance approval via Executive Vote
Result Full Prime with SubProxy, Artifact, Foundation, Token

What Gets Created:

  • Agent identity in the Synome
  • SubProxy Account (on-chain treasury)
  • Prime Agent Artifact
  • Foundation structure (legal entity)
  • Agent Token (10B genesis supply)

Current Prime Agents

Prime SubProxy addresses live in Appendix G (appendix-g-infrastructure.md).

Star Primes (5):

Prime Focus
Spark DeFi lending/liquidity
Grove Institutional credit, RWAs
Keel Solana ecosystem expansion
Star4 TBD
Star5 TBD

Institutional Primes (1):

Prime Focus
Obex Agent incubator, Prime + Halo development

Prime Capital Deployment

Allocation System

Deploy capital into approved strategies via PAU pattern.

Property Value
Pattern PAU (Parallelized Allocation Unit)
Constraint Rate limits on all operations

How It Works:

  1. Prime governance approves strategies via Init System
  2. GovOps instantiates approved configurations
  3. Sentinels execute operations within rate limits
  4. Capital flows through Controller → ALMProxy → Target

PAU (Parallelized Allocation Unit)

Standard building block for all capital deployment.

Component Purpose
Controller Entry point; enforces rate limits; validates operations
ALMProxy Holds custody; executes calls via doCall()
RateLimits Linear replenishment with configurable max and slope

Key Principle: Same contracts, different configuration. All layers use identical PAU patterns with layer-specific parameters.


Rate Limit System

Bounded capital movement velocity with linear replenishment.

Property Value
Pattern Linear refill from 0 to max
Parameters maxAmount (ceiling), slope (refill rate per second)
Consumption Decreases available limit; reverts if exceeds available

Rate Limit Types (per chain):

Limit Description
LIMIT_USDS_MINT Maximum USDS mintable
LIMIT_USDS_BURN Maximum USDS burnable
LIMIT_USDS_TO_USDC PSM swap limits
LIMIT_USDC_TO_CCTP Cross-chain transfer caps
LIMIT_USDC_TO_DOMAIN Per-destination bridging limits

SORL (Second-Order Rate Limit)

Constraint on rate limit increase speed.

Property Value
Max Increase 25% per cooldown period
Cooldown 18 hours between increases
Decrease Always instant (no constraint)

Purpose:

  • Prevent sudden capacity expansion attacks
  • Enable emergency decreases without delay
  • Give monitoring time to detect anomalies

Configurator Unit

Governance layer that controls rate limits and static parameters for PAUs without modifying deployed contracts.

Component Purpose
BEAMTimeLock Enforces 14-day delay on all additions (new PAU registrations, new inits, new role grants)
BEAMState Stores approved init configurations and accordant mappings
Configurator Operational interface where cBEAMs set rate limits and execute approved actions

BEAM Authority Cascade:

Level Role Holder Capabilities
1 Council Beacon (HPHA) Core Council (set by SpellCore) Root of all operational authority — sets aBEAMs, overrides all BEAM ownership, sets Synome write rights
2 aBEAM (Admin BEAM) Core Council Register PAUs, approve inits, grant cBEAMs — all via timelock. Instant removals for emergency response
3 cBEAM (Configurator BEAM) GovOps teams Set rate limits (within SORL), execute approved controller actions, manage relayer/freezer addresses
4 pBEAM (Process BEAM) Relayers / operational agents Direct execution on PAUs — calls Controller functions, moves capital within rate limits

Post-transition, the Council Beacon is set by SpellCore spells (16/24 Core Council Guardian hat) and is the single point from which all downstream BEAM authority flows. See governance-transition/council-beam-authority.md for the full authority model.

A GovOps team becomes accordant to a PAU when Core Council grants them the cBEAM for that PAU. The accordant GovOps has unified operational control: rate limit configuration, SORL increases, relayer management, and onboarding approved allocation targets.

Safety Asymmetry:

  • Additions (new PAUs, new inits, new permissions) — always timelocked (14-day delay), giving wardens time to review
  • Removals (revoke permissions, delete inits, unregister PAUs) — always instant, enabling rapid emergency response

Deployment Scope:

Each Configurator Unit manages PAUs within its governance scope:

Configurator Manages
Generator Configurator All Generator PAUs (single, central instance)
Mainnet Configurator All mainnet Prime + Halo PAUs
Altchain Configurator (per chain) All Foreign Prime + Foreign Halo PAUs on that chain

The Generator Configurator is separate from the Prime/Halo Configurators because Generators have a distinct governance scope (see architecture-overview.md). Prime and Halo PAUs on each blockchain share a single Configurator for that chain.

For detailed contract interfaces, invariants, and user stories, see smart-contracts/configurator-unit.md. For the post-transition Council Beacon authority model, see governance-transition/council-beam-authority.md.


Prime Risk Capital

Internal JRC (IJRC)

Prime's own junior risk capital; foundation of risk capacity.

Property Value
Source Prime's own treasury (SubProxy)
Position First-loss for Prime's exposures
Purpose "Skin in the game" for Prime

First Loss Capital (FLC):

Property Value
Size 10% of total JRC, sourced from Prime's own treasury (IJRC)
Position Absolute first-loss — absorbs before any other capital
Purpose Ensures Prime has direct skin in the game before external capital is touched

TEJRC Tokenization

External parties provide JRC via smart contract.

Property Value
Token Type TEJRC (Tokenized External JRC)
Standard LCTS-based (queue settlement)
Risk Junior position; higher yield compensates
Yield Set by Prime via dynamic formula matching supply/demand
Queue Rates Subscribe/redeem rates also set by Prime dynamically

Note: Yield and queue rate formulas are currently experimental and will be standardized over time.

Flow:

  1. External party deposits capital to TEJRC queue
  2. Capital converts at settlement based on available capacity
  3. TEJRC tokens represent junior risk capital position
  4. Earns yield; absorbs losses after First Loss Capital is depleted

TISRC Tokenization

Tokenized isolated senior risk capital per Prime.

Property Value
Token Type TISRC (Tokenized Isolated SRC)
Scope Isolated to specific Prime's exposures
Standard LCTS-based (queue settlement)
Risk Senior to JRC; absorbs only after JRC and agent token exhausted
Yield Set by Prime via dynamic formula matching supply/demand

Note: Yield and queue rate formulas are currently experimental and will be standardized over time.


Loss Absorption Waterfall

When losses occur, capital is consumed in strict order across three tiers:

Order Capital Layer Tier Mechanism
1 First Loss Capital (10% from IJRC) Prime First loss — absorbed entirely before moving to next layer
2 Remaining IJRC + EJRC Prime Split proportionally between Internal and External JRC
3 Agent Token Inflation Prime Dilute Prime token holders (potentially to infinity)
4 SRC Pool System TISRC merges into Global SRC (srUSDS); losses shared pari passu
5 SKY Token Inflation System Dilute protocol token holders
6 Genesis Capital Haircut Nuclear Protocol reserves
7 USDS Peg Adjustment Nuclear Final backstop — affects all USDS holders
Loss Event
    ↓
┌─── PRIME-LEVEL (per-Prime) ─────────────────────────────────────┐
│  1. First Loss Capital (10%)        │  ← Absolute first loss    │
├─────────────────────────────────────┤                           │
│  2. IJRC + EJRC (pro-rata)          │  ← Remaining junior       │
├─────────────────────────────────────┤                           │
│  3. Agent Token Inflation           │  ← Dilute Prime token     │
└─────────────────────────────────────────────────────────────────┘
                              ↓
┌─── SYSTEM-LEVEL (shared across all Primes) ─────────────────────┐
│  4. SRC Pool (pari passu)           │  ← TISRC + Global SRC     │
├─────────────────────────────────────┤                           │
│  5. SKY Token Inflation             │  ← Dilute SKY holders     │
└─────────────────────────────────────────────────────────────────┘
                              ↓
┌─── NUCLEAR OPTIONS (protocol reserves / peg) ───────────────────┐
│  6. Genesis Capital Haircut         │  ← Protocol reserves      │
├─────────────────────────────────────┤                           │
│  7. USDS Peg Adjustment             │  ← Final backstop         │
└─────────────────────────────────────────────────────────────────┘

Prime-Level (Steps 1-3): Losses are first absorbed by the specific Prime's capital stack. Agent Token inflation at step 3 can theoretically cover unlimited losses through dilution of Prime token holders.

System-Level (Steps 4-5): If Prime-level capital is exhausted, the insolvent Prime's TISRC merges into the Global SRC pool and losses are shared pari passu across all SRC holders. Sky charges a fee on TISRC yield for this protection. SKY inflation dilutes protocol token holders.

Nuclear Options (Steps 6-7): Genesis Capital and peg adjustment are last-resort mechanisms that should never be reached under normal conditions. These affect the entire protocol and all USDS holders.


Prime Rewards & Incentives

Distribution Rewards

Token rewards to incentivize USDS/sUSDS adoption through tagging. Rewards are tiered based on how strongly the distribution channel builds long-term demand-side moat.

Property Value
Recipient Primes (who then distribute at their discretion)
Split No enforced split — each Prime determines how to share with integrators
Eligible Assets USDS and sUSDS (both count as adoption)
Minimum Balance None

Distribution Reward Tiers:

Tier Rate Criteria
0 No DR Excluded/ineligible addresses
1 0 bps Untagged addresses (tracked but unpaid; can still earn Liability Duration Rewards)
2 10 bps Unbranded complex products (<90% sUSDS backing)
3 20 bps Branded USDS products OR unbranded with ≥90% sUSDS backing
4 50 bps Direct USDS/sUSDS holding with clear Sky branding (Boosted DR)

Tier 2 — Unbranded Complex Products (10 bps):

  • Product does not use "USDS" in name
  • Backed by mixed collateral with <90% sUSDS
  • Example: "StarUSD" containing various stablecoins as collateral

Tier 3 — Branded or High-Backing Products (20 bps):

Two paths qualify for Tier 3:

Path Requirements Example
3a: Branded Uses "USDS" in product name AND subscribe/redeem currency is USDS "StarUSDS"
3b: High Backing Unbranded but ≥90% backed by sUSDS "StarUSDT" with 90%+ sUSDS backing

Tier 4 — Boosted Distribution Rewards (50 bps):

  • Direct holding of USDS or sUSDS
  • Clear Sky branding in the frontend
  • Creates strongest demand-side moat

Tagging Mechanism:

Tagging associates a USDS/sUSDS balance with a Prime for Distribution Reward purposes.

Method How It Works
On-chain Codes embedded into transactions by frontends or smart contracts
Off-chain Verified by Guardian and GovOps

Tag Ownership Rules:

Property Value
Duration 10 years from tagging
Retagging Whoever tagged last owns the entire account
Transferability Tags cannot be transferred or sold

Note: Tag ownership rules (last-tagger-wins) are not yet finalized and may change.


Liability Duration Rewards

Rewards for Primes that source tagged USDS demand feeding into the ALDM (Asset-Liability Duration Matching) system.

Property Value
Purpose Compensate Primes for bringing sticky USDS demand (liability duration)
Eligibility Tagged USDS that sits in Duration Buckets
Split 2/3 to tagging Prime, 1/3 to Sky (initial setting)

How It Works:

  1. USDS accounts accrue liability duration over time and are placed in Duration Buckets based on their duration profile
  2. Primes tag USDS accounts, giving them a proportional "share" of each bucket's underlying demand
  3. When a Prime pays fees to reserve capacity, their reservation "tugs" on buckets to match their duration needs
  4. Fees flow back to whoever tagged the USDS that actually gets tugged

Dynamic Matching:

Bucket contents change over time as USDS accounts accrue or lose duration. When a Prime reserves capacity:

  • They may tug on nearby buckets if their target bucket lacks sufficient capacity
  • Fee distribution follows the actual tugging, not the original reservation target
  • This ensures rewards always flow to Primes whose tagged demand is actually being matched

Fee Distribution:

For each bucket that gets tugged, fees are distributed proportionally:

  • If Prime A tagged 30% of a bucket's USDS, Prime A receives 30% of the fees from tugs on that bucket
  • The split multiplier (initially 2/3) is then applied, with the remainder going to Sky

Example:

  • Prime B reserves capacity targeting Bucket 42, but tugs Buckets 41-43 based on availability
  • Bucket 42 contains $100M of USDS; Prime A has tagged $30M (30%)
  • Of the $1M fees Prime B pays, $400K is attributed to Bucket 42
  • Prime A receives: 30% × (2/3) × $400K = $80K from Bucket 42
  • (Plus any share from Buckets 41 and 43 if Prime A tagged demand there)

Even Tier 1 (0 bps DR) tagged addresses can earn Liability Duration Rewards, allowing Primes to benefit from demand-side contributions without requiring full branding or composition requirements.


Integration Boost

Mechanism to apply SSR yield to naked USDS held in smart contracts without using sUSDS.

Property Value
Rate Equivalent to Sky Savings Rate (SSR)
Applies To Naked USDS sitting in approved smart contracts
Purpose Enable USDS to earn yield in contexts where sUSDS integration isn't practical
Approval Per-integration; governance-approved

How It Works:

Integration Boost is essentially a manual way of having USDS earn the SSR without converting to sUSDS. This is useful for:

  • Protocols that need to hold raw USDS (e.g., for liquidity or specific contract logic)
  • Integrations where wrapping/unwrapping sUSDS adds friction or gas costs
  • Smart contracts that cannot easily integrate ERC-4626 vaults

The yield is paid out to the smart contract holding the USDS, which can then distribute it according to its own logic.


Governance Access Reward

Rewards for Primes providing compliant governance frontends.

Property Value
Pool 1% of Sky's yearly net revenue
Recipient Primes with compliant SKY Staking and governance frontends
Requirements Frontend meets Atlas and Synome specifications
Distribution Split among eligible Primes based on SKY staked through each frontend

Note: Specific compliance requirements for frontends are TBD.


Prime: Duration Capacity Reservation (Planned)

Acquire duration-matching capacity via governance allocations initially, then via auctions once Prime-side stl-base is live.

Property Value
Purpose Reserve Duration Bucket capacity for long-duration assets
Mechanism Governance allocations (pre-auction), then daily sealed-bid auctions (once stl-base is live)
Benefit Lower capital requirements when assets match liability duration

101 Duration Buckets:

  • Each bucket = 15 days
  • Bucket 0 = immediate liquidity
  • Bucket 84 = 1,260 days (JAAA CLO AAA)
  • Bucket 100 = 1,500+ days (structural/permanent)

Prime: Pioneer System (Genesis Stars Only)

The Pioneer System enables Stars to gain exclusive advantages when expanding USDS to new blockchains. This primitive is only available to the 5 genesis Stars.

Property Value
Availability Genesis Stars only (Spark, Grove, Keel, Star4, Star5)
Purpose Incentivize cross-chain expansion with first-mover advantages
Duration 3-year Pioneer Phase per Pioneer Chain
Exclusivity One Pioneer Star per chain; Stars can have multiple Pioneer Chains

Pioneer Chains and Pioneer Stars

A Pioneer Chain is any blockchain integrating USDS for the first time. A Pioneer Star is the Star designated by the chain's team/foundation to lead the USDS rollout.

Designation Process:

  1. Written confirmation from Pioneer Chain's official team/foundation
  2. Star's governance verifies intention to accept Pioneer Status
  3. Core Council reviews strategic alignment with Sky Ecosystem growth
  4. If approved, Pioneer Phase (3 years of exclusive benefits) begins

Pioneer Phase Benefits

1. Distribution Reward Tagging

Property Value
During Pioneer Phase Pioneer Star auto-tags all untagged USDS/sUSDS accounts on the chain
At Phase End One-time permanent tag of remaining untagged balances
Tag Duration 10 years (unless retagged by another Star)
Boosted DR Not available on auto-tags; only on explicitly tagged + approved

2. Pioneer Rewards

Property Value
Source SSR × Unrewarded USDS balance on Pioneer Chain
Timing Paid each Monthly Settlement Cycle
Recipient Pioneer Star's SubProxy (as income)

"Unrewarded USDS" = bridged USDS not earning SSR through sUSDS, Integration Boost, or Agent holdings.

3. Unbalanced Supply Fee Authority

The Pioneer Star can charge fees to other Stars that allocate supply without contributing to demand.

Property Value
Fee Rate 20 bps per year
Applies To Unbalanced supply (collateral/liquidity from USDS debt not matched by demand)
Purpose Protect supply/demand balance on Pioneer Chain

Balancing Mechanisms (to avoid fee):

Mechanism Offset Ratio
Tagging Demand 1:1 (e.g., $1M tagged USDS offsets $1M supply)
ASC Liquidity 3× average daily trading volume counts as demand

Exemptions: Pioneer Star or Pioneer Chain team can grant exemptions via Ecosystem Accord. Exemption terms cannot be revoked if the exempted Star made significant investments based on them.

De-designation

If the Pioneer Star fails to meet expectations recorded in an Ecosystem Accord, the Pioneer Chain team/foundation can:

  • De-designate the Pioneer Star, or
  • Transfer designation to a different Star

When de-designated, the 3-year clock pauses. A future Pioneer Star continues from where the previous one stopped.


Halo Agents

Halo Agents are investment products created by Primes. They wrap specific strategies and deploy capital into real-world assets, custodians, and regulated endpoints.

Halo Classification

Halos are classified on two dimensions: regulatory treatment and Halo Class type (which determines the capital deployment mechanism).

By regulatory treatment:

Treatment Types Rules
Standard Portfolio, Term, Trading Normal Halo rules — standard governance, rate limits, and risk framework
Special Identity Network, Exchange Additional regulatory or operational requirements beyond standard Halo rules

Standard Halo Class types (by capital deployment mechanism):

Class Type Token Standard Use Case
Portfolio Halo LCTS (pooled, fungible) Standardized products with uniform terms — all participants share the same conditions
Term Halo NFAT (individual, non-fungible) Bespoke deals with negotiated terms — each position has different duration, size, and yield
Trading Halo AMM (programmatic counterparty) Instant liquidity for RWA tokens and ecosystem assets via automated market making

A Halo Agent with a single Halo Class is commonly referred to by its class name (e.g., "a Portfolio Halo" or "a Trading Halo").

Special Halos:

Type Purpose
Identity Network Halo Operates identity verification infrastructure (KYC registries). See sky-agents/halo-agents/identity-network.md
Exchange Halo Operates intent-based exchange infrastructure (orderbooks, matching engines). See trading/sky-intents.md

Halo Creation

Property Value
Creator Parent Prime only
Method Factory deployment from templates
Result Halo with Unit Structure, Artifact, token standard integration

Halo Class, Book, and Unit Structure

Halos organize capital into three layers: Class, Book, and Unit. Each layer serves a distinct purpose — infrastructure sharing, balanced ledgers, and cross-book links.

Hierarchy:

Halo (Synomic Agent)
└── Halo Class (shared SC + legal infra)
    ├── Halo Books (balanced ledgers: assets = liabilities)
    └── Halo Units (cross-book links connecting layers)

Halo Class — shared smart contract infrastructure (PAU, LPHA beacon) and legal framework (buybox, agreements). The Class defines operational bounds.

What a Class shares What varies per Unit
PAU (Controller + ALMProxy + RateLimits) Tranche seniority (for Portfolio)
LPHA beacon (lpha-lcts, lpha-nfat, or lpha-amm) Duration, size, specific terms (for Term)
Legal framework (buybox constraints) Asset pair, spread, pool depth (for Trading)
Factory template (audited, reusable) Risk/return profile within class constraints

Halo Book — a balanced ledger (assets = liabilities) that serves as a bankruptcy-remote isolation boundary. Each book balances real-world positions on the asset side against the Units that claim on them on the liability side. Books provide risk isolation: units sharing a book are pari passu on losses; units on different books are fully isolated. Multiple assets can be blended in a book for borrower privacy. Books progress through a lifecycle (Filling → Deploying → At Rest → Unwinding) with capital requirements varying by phase — see sky-agents/halo-agents/halo-class-book-unit.md.

Halo Unit — a cross-book link representing a claim on a specific Book. Each Unit is legally and operationally isolated from other units (similar to a serialized LLC). The token standard determines how claims are represented:

Standard Model Use Case
LCTS Pooled, fungible shares Many users, same terms, shared capacity (ETF-like)
NFATS Individual, non-fungible tokens Bespoke deals, named counterparties (bond-like)

Examples:

Class Type Halo Class Example Halo Units Within
Portfolio Tranched CLO structure Senior tranche, Junior tranche (same PAU, same lpha-lcts)
Term NFAT Facility with buybox Multiple NFAT deals (varying duration, size within buybox)
Trading RWA instant-settlement pool USDS/T-Bill pair, USDS/JAAA pair (same AMM, same lpha-amm)

The underlying smart contract infrastructure (PAU + LPHA beacon) remains consistent across a Halo Class; the token standard determines the user-facing mechanics for subscribing, redeeming, and transferring positions.


Generator Agents

Generator Agents issue Sky Generated Assets (SGAs) — stablecoins for currencies beyond USD. Currently, there is one implicit Generator for USDS; future Generators will issue other currencies.

Note: Generator creation process and detailed specifications are TBD.

Generators have a limited set of specific primitives available to them:

Issue Generated Asset

Property Value
Types Sky Generated Asset (SGA) or External Generated Asset (EGA)
Examples USDS (current), future: additional currency SGAs
Mechanism Mint new stablecoins backed by collateral

Issue Global Senior Risk Token

Property Value
Type srSGA (e.g., srUSDS)
Purpose Global senior risk capital for the generated asset
Standard LCTS-based (queue settlement)

Issue Fixed Rate Tokens

Property Value
Types Fixed-rate sSGA, Fixed-rate srSGA
Purpose Provide predictable yield versions of savings and risk tokens
Mechanism Lock in rate at issuance

Set Distribution Rewards

Property Value
Purpose Incentivize adoption of the generated asset
Mechanism Configure reward rates for integrations

Set Sticky Demand Rewards

Property Value
Purpose Distribute duration auction income to Primes (once auctions are live)
Recipients Primes that have sourced sticky demand (long-duration liabilities)
Source Revenue from Duration Bucket capacity auctions (once auctions are live)

Guardian Agents

Guardian Agents operate systems with collateral backing. They interpret governance documents, participate in governance, execute day-to-day operations, and provide oversight within the Sky governance framework. Guardians consolidate the former Facilitator (interpretation), Aligned Delegate (governance participation), and Executor (operations) roles into a single collateral-backed role.

Guardian Types

Type Role Accountability
Operational Guardian Day-to-day execution Posts collateral; subject to derecognition
Core Guardian Interprets Atlas, oversees Operational Guardians, participates in governance Atlas alignment; collateral-backed interpretation

Operational Guardian

Property Value
Function Execute routine operations within defined bounds
Collateral Posts insurance backing their operations
Supervision Subject to Core Guardian oversight
Risk Derecognition and collateral loss for violations

Core Guardian

Property Value
Function Interpret Atlas, oversee Operational Guardians, participate in governance votes
Relationship Governs multiple Operational Guardians; receives delegated SKY voting power
Accountability To Sky Governance; collateral-backed interpretation and execution

Folio Agents

Folio Agents are rank 3 standardized supply-side holding structures in the Sky ecosystem. They are not Halos — the structural difference is fundamental: a Halo wraps around a legal entity, while a folio is controlled BY the principal through legal entities (the principal's legal structure is the governance surface).

Each folio has a principal (the end user) who controls the folio through a directive — human language instructions defining investment philosophy, risk appetite, and constraints. Two operating modes:

  • Automated folio — operated by a sentinel formation (baseline + stream + wardens) via guardian accord. The directive governs the formation's behavior.
  • Principal control folio — operated directly by the principal via a principal sentinel (stl-principal). No guardian accord, no formation. The principal deploys their own algorithms and manages positions directly.
Property Value
Rank 3 — administered by a Prime
Token None — tokenless, single owner (the principal)
Creation Instant — any SKY holder
Purpose Standardized holding structure: Growth Staking, capital deployment, strategy access
Control Automated (sentinel formation) or Principal Control (principal sentinel)
Restructuring Type 1 only (graduate to Halo or Prime)

See sky-agents/folio-agents/agent-type-folios.md for the full specification.


Core Controlled Agents

Core Controlled Agents are rank 1 agents directly administered by the Core Council. They serve as general-purpose operational vehicles — in the short term covering legacy protocol positions (replacing "Core Halos"), in the long term serving any Core Council operational need.

Property Value
Rank 1 — directly regulated by Core Council
Token None — tokenless
Creation Core Council governance (SpellCore)
Purpose Legacy asset management, general-purpose Core Council operations
Growth asset No — excluded from Growth Staking

See sky-agents/core-controlled-agents/agent-type-core-controlled.md.


Recovery Agents

Recovery Agents are rank 1 crisis agents administered by the Core Council, activated when a Guardian collapses or is implicated in misconduct.

Property Value
Rank 1 — directly regulated by Core Council
Token None — tokenless
Creation Core Council governance (SpellCore), crisis response
Purpose Take over affected agent tree, manage resolution
Duration Temporary — dissolves after crisis resolution
Growth asset No — excluded from Growth Staking

See sky-agents/recovery-agents/agent-type-recovery.md.


This appendix should be updated as new primitives are deployed.