Spark Savings

Protocol identification — as of October 3, 2026

• Name / website / docs: Spark Savings; https://spark.finance/; documentation at Spark Docs. Category: Yield / tokenized savings vaults. DeFiLlama ID 6716 identifies it as “Spark Savings.”
• Launch: Spark Protocol launched May 9, 2023; the first Spark Savings product launched in August 2023. Spark Savings V2 is a later ERC-4626 product; its audited code assessment is dated October 1, 2025.
• Chains: Requested scope—Ethereum, Arbitrum, Avalanche, Base, and Robinhood Chain. Current documentation additionally lists Gnosis, Optimism, Unichain, and X Layer. Base is confirmed as a Spark Liquidity Layer deployment, but a Base V2 vault address was not located in the reviewed address tables.
• Native token: SPK. Vault receipt tokens are separate assets, including spUSDC, spUSDT, spETH, spPYUSD, and spUSDG. Representative main V2 vault addresses
• Ethereum: spUSDC 0x28B3…a43d; spUSDT 0xe2e7…c372; spETH 0xfE6e…9B8f; spPYUSD 0x8012…D354.
• Avalanche: spUSDC 0x28B3…a43d.
• Arbitrum: spUSDT 0x45d9…4b62.
• Robinhood Chain: spUSDG 0xde77…4087. Address verification: Explorer evidence confirms Ethereum spUSDC as a deployed, source-labelled Spark vault; Avalanche and Robinhood explorer pages were located. However, the required two-source cross-check including Dune cannot be completed because Dune MCP is unavailable: Not verifiable as of October 3, 2026. Base V2 address: Not verifiable as of October 3, 2026. Fork lineage Spark Vaults V2 is explicitly a fork of Sky’s sUSDS design. Changes include removing DSS hooks, replacing wards/rely/deny with OpenZeppelin role-based access control, adding SETTER_ROLE, TAKER_ROLE, take(), assetsOutstanding(), referrals, and UUPS upgradeability. The V2 implementation was audited by ChainSecurity; the assessment reported zero critical/high/medium findings and one low finding corrected. Similar-fork malicious-modification history: Not verifiable as of October 3, 2026.

Spark Savings appears to be a real, functioning product rather than a pure landing page: the docs point to an app at app.spark.finance/savings, and the user guides describe actual deposit and withdraw flows, vault APY, TVL, users, deposit caps, and available liquidity shown in the UI. The integrator docs also expose a read-only Savings Data API plus a machine-readable OpenAPI spec at api.spark.finance/openapi.json, so there is an open API for product data. I did not find evidence of obvious broken links, fake metrics, or template-site behavior in the current sources; instead, the documentation is internally consistent and repeatedly references the live app and vault workflows.

The protocol site and docs also warn about scams and fake accounts, which is a standard security notice rather than a sign of a fake portal. ## Maturity assessment

• Real portal: yes, the docs and product pages describe a live app experience.
• Live deposits/withdrawals: yes, documented end-user flows exist and are presented as operational.
• Docs/UX: mature enough to include guided tutorials, vault details, and developer integration docs.
• Open API: yes, a read-only Savings Data API with OpenAPI spec is published.
• Broken links / fake metrics / template signs: not verifiable as of 2026-10-03; no clear evidence was found in the available sources.

Spark has an active bug bounty program managed by Immunefi. It went live on 2023-11-01. The program’s critical smart-contract bounty is 10% of funds directly affected, capped at USD 5,000,000, with a minimum critical reward of USD 50,000.

Web/app rewards are capped at USD 50,000 maximum / USD 5,000 minimum. The program requires a PoC for critical and high smart-contract reports, and Immunefi states payouts are denominated in USD but paid in DAI at 1:1 parity. Public program pages also state it is triaged by Immunefi and that bounty payouts are handled by MakerDAO governance.

Publicly available pages did not surface disclosed incident counts or payout totals, so results are not verifiable as of 2026-10-03.

Assessment date: October 5, 2026. Dune is unavailable; on-chain verification, exact wallet balances, and contract-level exposure percentages were skipped. Dependency and counterparty assessment

• Primary dependency: Spark Savings V2 routes assets through the Spark Liquidity Layer; legacy/Sky vaults depend on Sky Governance and the Sky Savings Rate. The documentation states allocations span DeFi and RWA strategies, but does not provide a complete, independently verifiable allocation schedule. This is an unverified marketing claim where sourced only from Spark.
• External protocols/RWAs: Spark publicly names or references BUIDL, Superstate, Maple/SyrupUSDC, and Centrifuge exposures. Issuer, SPV, custodian, redemption, maturity, and concentration details are Not verifiable as of October 5, 2026. Failure, default, legal seizure, NAV impairment, or liquidity gates at an RWA issuer/custodian could impair withdrawals or USDS backing.
• Oracle/manipulation risk: Sky documents explicit reliance on third-party oracles and keepers. Incorrect, delayed, or manipulated prices could cause over-borrowing, bad collateral valuation, unnecessary liquidations, or insolvency. DEX, solver, RPC, and smart-contract dependencies add correlated failure paths.
• Bridges and chain risk: Spark is shown by DeFiLlama on eight chains, while the supplied scope lists five; this is a scope discrepancy. Governance records show cross-chain USDT routes to Arbitrum and X Layer. Bridge compromise, sequencer/chain outage, message failure, or liquidity fragmentation could block redemptions. Exact bridge exposure is Not verifiable as of October 5, 2026.
• Stablecoin/LST/restaking: Supported assets include USDC, USDT, PYUSD, USDS, USDG, and ETH. stUSDS is specifically described as higher risk because of liquidity constraints and lack of automatic liquidations. LST/restaking and exact stablecoin concentration are Not verifiable as of October 5, 2026.
• Custody/CEX/MM: No direct CEX, market-maker, or third-party custody concentration is disclosed. Not verifiable as of October 5, 2026.
• Failure scenario: A severe loss could exhaust junior capital/backstops and be socialized across USDS holders; Spark may enter recovery mode and temporarily halt withdrawals. USDS depeg, RWA impairment, oracle failure, bridge exploit, or Sky governance error are material tail risks. Contradiction / data gap: DeFiLlama reports approximately $1.321B TVL, but no raw on-chain verification was possible and the chain list differs from the supplied scope. The on-chain number is therefore Not verifiable as of October 5, 2026. No active dependency failure was identified in the reviewed sources as of October 5, 2026.

Spark Savings is organized as a non-custodial vault system: users deposit stablecoins into permissionless ERC-4626 vaults and receive vault shares/tokens that represent their claim on the underlying assets, so the protocol’s contracts coordinate liquidity and redemptions rather than acting as a custodian. The public docs also describe larger withdrawals being handled via liquidity buffers and, when needed, asynchronous liquidity intents; withdrawals are presented as available at any time, with no lockups. There is also an institutional distribution path where assets may remain in regulated custody with partners such as Anchorage or BitGo while being deployed into Spark markets, but that is an institutional access arrangement rather than the core retail vault custody model.

As of October 3, 2026, Spark Savings uses role-based smart-contract permissions, not a publicly documented user/operator threshold-signature scheme.

• User control: Savings positions are ERC-4626 vault shares held by the user’s wallet. Users approve withdrawals or intent contracts to spend shares; the Savings Intents contract is documented as not holding user assets or shares.
• Vault authority: Spark Savings V2 contracts use OpenZeppelin AccessControl and are upgradeable. The documented roles are:
• DEFAULT_ADMIN_ROLE: can upgrade implementations and set key vault parameters;
• SETTER_ROLE: can update the vault savings rate within configured bounds;
• TAKER_ROLE: assigned to the Spark Liquidity Layer and can withdraw liquidity from the vault for deployment into yield strategies.
• Withdrawal operations: Large withdrawals may be fulfilled by the off-chain ALM Planner through the RELAYER role. The relayer can execute a user’s already-created withdrawal request, but the documented design says it cannot arbitrarily transfer vault assets; the request is redeemed against the user’s share allowance and sent to the specified recipient.
• Governance layer: Spark Governance controls protocol parameters including risk settings, integrations, and deployments; the exact signing quorum, signer identities, key-storage technology, and rotation procedures are not disclosed in the reviewed sources. Not verifiable as of October 3, 2026.
• Operational resilience: A permissionless-withdrawal mechanism is documented as a backup path that can let users trigger certain withdrawals without the ALM Planner being online, subject to supported assets and conditions. Important correction to the prior finding: Spark’s published threshold-signature description concerns the separate Bitcoin/off-chain Spark architecture, where users and operators hold complementary key shares. It should not be treated as evidence that Spark Savings’ EVM vault administrator, taker, relayer, or upgrade keys use threshold signatures. Risk conclusion: Spark Savings is user-controlled at the share-ownership layer, but protocol-level liquidity deployment, upgrades, rate changes, and some withdrawals depend on privileged operational keys. The precise custody architecture for those privileged keys is Not verifiable as of October 3, 2026.

Scope/as-of: October 5, 2026. Dune MCP was unavailable, so deployment-specific addresses, proxy-admin classification, decoded admin events, current role holders, renouncement, timelock delay, and chain-by-chain deployment verification are Not verifiable as of October 5, 2026. This applies separately to Arbitrum, Avalanche, Base, Ethereum, and Robinhood Chain; no exposure percentages are asserted. Verified design evidence (not deployment-state evidence): Spark Vaults V2 are ERC-1967/UUPS proxies. DEFAULT_ADMIN_ROLE can upgrade implementations, assign/manage roles, and set VSR bounds; SETTER_ROLE changes VSR within bounds; TAKER_ROLE calls take() and is documented by ChainSecurity as able to fully drain vault-held liquidity. Therefore the design-level admin_can_drain assessment is true, although the deployed holder and whether governance controls it are unverified. Controls and exit: ERC-4626 deposit/mint/withdraw/redeem are present.

Withdrawals can be constrained by idle liquidity; PSM conversion fees or disabled conversion can make USDC exits revert, while an sUSDS escape hatch may remain available. Permissionless Withdrawals provides a backup path for supported assets, sourcing liquidity through the ALM Controller/ALMProxy, subject to whitelist, coverage and penalty limits. Worst case if privileged keys are compromised: upgrade the proxy to malicious logic; grant roles; set harmful VSR/caps; use take() to remove vault liquidity; alter withdrawal configuration or connected liquidity infrastructure. Result: theft, insolvency, frozen withdrawals, or forced adverse exits.

No verified on-chain timelock or emergency-pause guarantee exists. Architecture: User → ERC-4626 Spark Vault proxy → underlying asset / sUSDS or PSM3 → ALM Controller + ALMProxy → whitelisted venues/strategies. Admin/governance → proxy upgrade + AccessControl; TAKER → liquidity deployment; permissionless-withdrawal module → user redemption. Audit: ChainSecurity’s October 1, 2025 source-code assessment reported 0 open critical/high findings, but it covered source code—not deployed addresses or current configuration. Addresses and deployment-wide verification: Not verifiable as of October 5, 2026.

Report: Spark User Actions; auditor: ChainSecurity; publication date: 2024-09-04; scope: PSMVariant1Actions and MigrationActions; link: https://reports.chainsecurity.com/Spark/ChainSecurity_Spark_SparkUserActions_Audit.pdf; covers deployed code: Not verifiable as of 2026-09-04.

Report: Savings USDS; auditor: ChainSecurity; publication date: 2024-09-30; scope: SUsds/ISUsds and deployment scripts, including L2 SUsds; link: https://reports.chainsecurity.com/MakerDAO/ChainSecurity_MakerDAO_SavingsUSDS_Audit.pdf; covers deployed code: Not verifiable as of 2026-09-04 (bytecode match not performed).

Report: Spark Vaults; auditor: ChainSecurity; publication date: 2025-01-24; scope: USDC ERC-4626 vault and L2 deployment; link: https://www.chainsecurity.com/reports/Spark/ChainSecurity_Spark_SparkVaults_Audit.pdf; covers deployed code: Not verifiable as of 2026-09-04.

Report: Spark Vaults V2; auditor: ChainSecurity; publication date: 2025-10-01; scope: SparkVault ERC-4626 contract; link: https://www.chainsecurity.com/reports/Spark/ChainSecurity_Spark_SparkVaultsV2_Audit.pdf; covers deployed code: Not verifiable as of 2026-09-04.

Report: Spark Savings Intents; auditor: ChainSecurity; publication date: 2026-03-11; scope: SavingsVaultIntents.sol and interfaces; link: https://reports.chainsecurity.com/Spark/ChainSecurity_SparkDAO_SparkSavingsIntents_Audit.pdf; covers deployed code: Not verifiable as of 2026-09-04.

Spark DAO – Spark Savings Intents / Savings Vaults (including boosted vaults) smart‑contract audits.

Boosted Spark Savings Vault audit.

Report: Permissionless SLL Withdrawals; auditor: ChainSecurity; publication date: 2026-08-17; scope: permissionless withdrawal contracts; link: https://reports.chainsecurity.com/Spark/ChainSecurity_SparkDAO_PermissionlessSLLWithdrawals_Audit.pdf; covers deployed code: Not verifiable as of 2026-09-04.

Assessment: Spark Savings appears to be a real, publicly attributed DeFi business rather than an anonymous web front. It is developed by Phoenix Labs within the Sky ecosystem (formerly MakerDAO), but the legal and operating structure is contributor/service-provider based rather than a conventional single operating company. Founders and team

• Sam MacPherson — co-founder and CEO of Phoenix Labs; previously worked in MakerDAO protocol engineering, at Maple, and at EY.
• Lucas Manuel — co-founder and head of smart contracts; associated with prior MakerDAO smart-contract engineering.
• Nadia Alvarez — identified as a co-founder/COO; previously MakerDAO growth and Latin America business development, and worked at Red Hat.
• Kris Kaczor — identified as CTO. Blockworks reports that key leadership identities are publicly disclosed and describes the broader Phoenix Labs team as non-anonymous. Prior projects and credibility The strongest credibility signal is continuity with MakerDAO/Sky: Spark launched in 2023 as a MakerDAO-linked lending product and later expanded toward savings, liquidity allocation, and deployment across DeFi, CeFi, and real-world assets. This indicates experienced incumbents rather than first-time founders, although prior MakerDAO affiliation is not itself evidence that every Spark deployment is safe. Hacks and outcomes A current independent record of Spark-specific exploits was not established in the available results. A third-party report states that no documented exploit occurred during its reviewed operating period, but this is not an exhaustive incident database and should not be treated as proof of absence. Claims about security scans and “no critical vulnerabilities” come from a security-provider interview and are not equivalent to an independent incident or audit history. Public vs. anonymous: Public leadership; no evidence here of anonymous founders. Office and jurisdiction: Sam MacPherson’s public profile places him in Toronto, Canada. A verified Phoenix Labs registered office, domicile, incorporation details, or onshore/offshore classification was not established from the available independent evidence: Not verifiable as of 2026-10-03. Reality-check finding: Phoenix Labs is described as a private company providing technical and operational services to the Spark Foundation, while Spark is governed within the Sky/MakerDAO ecosystem. Therefore, “Spark” should not automatically be treated as a conventional incorporated fintech with a single balance sheet, headquarters, or parent-company liability. The protocol’s institutional credibility is materially stronger than its publicly verified corporate-footprint transparency.

Spark Savings has a moderately positive but not low-risk reputation. It is generally viewed as a legitimate component of the Sky ecosystem, with experienced upstream governance, multiple security reviews, and no publicly documented major exploit or insolvency event identified in the reviewed material. These findings are based on web sources only; on-chain verification was unavailable, so TVL, exposure, ownership, and contract-state claims are Not verifiable as of 2026-10-03. Protocol, founders, and backing

• Spark is closely tied to Sky, the successor ecosystem to MakerDAO, rather than being an anonymous standalone project. This provides operating history and institutional visibility, but also creates dependency on Sky governance and its core decision-makers.
• Public risk assessments characterize the main concern as structural and governance-related, not evidence of a rug pull or predatory scheme.
• Investor identities, capitalization, and founder-level financial interests are Not verifiable as of 2026-10-03 from the reviewed independent sources. Audits and security
• Security coverage is comparatively strong: ChainSecurity and other firms have reviewed Spark-related lending, savings, and liquidity components, and Sky maintains an Immunefi bounty with a stated maximum payout of $5 million.
• A documented oracle concern involved possible inaccurate pricing during wstETH, rETH, or weETH depegs; the reviewed account says governance subsequently upgraded the relevant oracle design.
• No publicly documented major exploit or paid bounty was identified in the reviewed sources, but audit coverage does not eliminate implementation, upgrade, oracle, or operational risk. Key criticisms and unresolved concerns
• Governance concentration: Sky controls important Spark parameters and execution pathways; critics question the degree of effective decentralization and the ability of token holders to constrain upstream governance.
• Systemic dependency: A Sky/USDS failure, depeg, bad debt event, regulatory action, or adverse governance decision could impair savings redemptions and yields.
• Counterparty and RWA exposure: Spark’s liquidity-routing model can transmit risks from external protocols and Sky’s real-world-asset strategy.
• Token concentration and unlock pressure: Independent assessments flag concentrated SPK allocations and potential dilution or cliff-unlock risk.
• Operational opacity: Some assessments note undisclosed multisig signers and contracts or execution components that are difficult to independently verify. Legal, regulatory, and sanctions
• The reviewed material identifies regulatory ambiguity around yield-bearing stablecoins potentially resembling securities or regulated deposits, while frontends use geographic and sanctions-related access controls.
• Specific enforcement action, sanctions designation, fraud finding, or insolvency proceeding against Spark Savings was Not verifiable as of 2026-10-03. Overall, reputation is legitimate and institutionally oriented, but dependent on centralized Sky governance, stablecoin/RWA counterparties, and regulatory treatment. No verified rug, fraud, or insolvency event was found in the available evidence.

As of October 5, 2026. Spark Savings is a yield-bearing vault suite: users deposit stablecoins or ETH and receive ERC-4626-style shares whose value accrues through underlying strategy returns. Stablecoin deposits are pooled across entry assets and routed through the Spark Liquidity Layer/SubDAOs into USDS collateral, on-chain lending, tokenized U.S. Treasuries, tokenized funds, private credit, stablecoin liquidity and OTC lending. Yield classification. This is primarily strategy-generated yield, not demonstrably liquidity-mining subsidized yield.

However, the organic share is Not verifiable as of October 5, 2026. Stablecoin vaults are USD/delta-light but not market-neutral in the risk sense: they retain USDS depeg, credit, counterparty, RWA and governance exposure. Savings ETH is directional ETH exposure; Staked USDS adds SKY-collateralized lending exposure.

No restaking was identified. Exact leverage/looping exposure is Not verifiable as of October 5, 2026. Liquidity mechanics. Stablecoin vaults retain an idle-liquidity buffer for withdrawals; Savings ETH reportedly holds approximately 5–10% idle. Share redemption is supported by the ERC-4626 vault design.

Specific withdrawal cooldowns, gates, per-wallet limits and all fee schedules are Not verifiable as of October 5, 2026. Spark’s “fee-free access” is an unverified marketing/aggregator claim, not independently confirmed here. TVL / APY. DeFiLlama reports $1.34B TVL, +9.1% over 30 days, and 3.34% average supply APY across eight pools. By chain: Ethereum $1.001B (~74.7%), Arbitrum $301.9M (~22.5%), Robinhood Chain $14.7M (~1.1%), Avalanche $8.6M (~0.6%), Base $5.9M (~0.4%); the remainder is on other chains.

Dune TVL, product-level attribution, historical APY volatility and Dune-vs-DeFiLlama reconciliation: Not verifiable as of October 5, 2026. Revenue/collateral. Savings-specific protocol revenue is Not verifiable as of October 5, 2026; DeFiLlama’s $1.06M 30-day revenue figure is for parent Spark, not Savings alone.

As of October 5, 2026: Size / contradiction. Spark’s Q2 2026 internal financial report states treasury capital of $48.5M at June 30, 2026, used as junior/first-loss capital. Spark’s live financial dashboard currently displays $43.72M Total Treasury. These figures are not directly reconciled; the dashboard is protocol-operated and the Q2 figure is unaudited. Composition. The Q2 report describes treasury returns primarily from USDS savings yield, with an unrealized loss on a token position; it does not provide a complete, independently verified asset-by-asset composition in the accessible text.

Treasury-related operating capital is distinct from user vault assets and SLL allocations. Spark Savings V2 assets are described as USDS-backed, while spETH is backed by Spark’s balance sheet and insurance. Known addresses / control. Key Ethereum addresses disclosed by Spark include Spark Treasury 0x3300f198988e4C9C63F75dF86De36421f06af8c4, SAF 0xEabCb8C0346Ac072437362f1692706BA5768A911, SparkLend collector 0xb137E7d16564c81ae2b0C8ee6B55De81dd46ECe5, Foundation 0x92e4629a4510AF5819d7D1601464C233599fF5ec, and operations safe 0x2E1b01adABB8D4981863394bEa23a1263CBaeDfC. Governance records show the SubDAO Proxy as vault default admin, the operations safe as setter, and chain-specific ALM proxies as takers. Custody / reserve policy. Reserves are held through on-chain treasury and ALM infrastructure; the report also references custody-supported institutional lending, including an Anchorage tri-party structure.

Policy language requires sufficient liquid reserves for expected withdrawals and positions treasury as junior capital absorbing losses before senior/system capital. Attestations and on-chain verification. No independent financial audit or reserve attestation was identified. The Q2 report explicitly describes internal financial records and unaudited estimates. Dune MCP is unavailable in this run; therefore balances, composition, and cross-chain exposure are Not verifiable as of October 5, 2026. Known liability. The Q2 report discloses $6.36M accrued depositor-yield liability at June 30, 2026; total liabilities are not disclosed. Assessment: reserve transparency is moderate but dependent on protocol reporting; independent verification, full composition, custody segregation, and current on-chain balances remain unresolved.

Status (as of October 3, 2026): Spark Savings is a Spark product, and it now has an ecosystem token: Spark (SPK). This corrects the September 3, 2026 finding that no native token existed. SPK is not a vault receipt token; sUSDS, spUSDC, spUSDT, spETH and spUSDG are yield-bearing Savings tokens.

• Contract: Ethereum 0xc20059e0317DE91738d13af027DfC4a50781b066; Base 0x692A07f2306a3bba739e5281A26A5a97C6D7A6cA. Official documentation does not list SPK deployments on Arbitrum, Avalanche, or Robinhood Chain: Not verifiable as of October 3, 2026.
• Supply/valuation: Genesis/max supply is 10 billion SPK. CoinGecko reports approximately 3.4 billion circulating, about $85.4 million market cap; implied FDV is approximately $251 million at the same price. These are aggregator figures, not on-chain-verified.
• Utility/governance: SPK supports governance voting/delegation, staking, and SPK farming rewards. Spark governance controls risk settings, budgets, integrations and deployments; proposal submission requires 1% of total supply.
• Revenue/buybacks/burns: No documented revenue share or buyback program found. The verified token ABI exposes burn, but no evidence was found that protocol revenue funds systematic burns. Staking rewards are described as Spark Points/Symbiotic Points; SPK farming distributes SPK.
• Allocations/unlocks: 65% Sky farming/users over 10 years; 23% ecosystem; 12% team. Team vesting: 12-month cliff, then three-year vesting. TGE was June 16, 2025; final scheduled vesting is approximately May 2035. Whether every announced unlock occurred on-chain: Not verifiable as of October 3, 2026 without Dune/live chain reconciliation.
• Admin risk: Contract exposes minting, burning, rely/deny, and file authorization functions. Current ward/admin identities and any blacklist or fee switch are Not verifiable as of October 3, 2026.
• Holder concentration, insider wallets, DEX depth/listings: Not verifiable as of October 3, 2026; Dune was unavailable for this run.

Assessment: A Bitcoin fall below $10,000 would be an indirect, potentially limited stress for Spark Savings—not a direct BTC-price loss mechanism—because Spark Savings deposits are represented in the available evidence as USD-denominated deposits backed 1:1 by USDS. The principal transmission channel is impairment or liquidation losses in SparkLend collateral and broader Sky/USDS backing, not BTC exposure inside the savings vault itself.

• BTC collateral risk: Spark documentation states that BTC e-mode was fully removed in June 2026 and remaining positions were force-liquidated on June 8, 2026. An independent analysis reported affected exposure of approximately $1.6 million from one major borrower plus smaller positions before removal. This materially reduces direct BTC-linked liquidation risk, although current residual exposure is Not verifiable as of October 10, 2026 because on-chain verification was unavailable.
• Liquidation mechanics: SparkLend positions become liquidatable below a health factor of 1; up to 50% of debt may be covered when the factor is between 0.95 and 1, and up to 100% at or below 0.95. Falling BTC prices could therefore create stress only where BTC-linked collateral remains accepted or indirectly affects correlated collateral and liquidity.
• Oracle and contagion controls: The documentation describes peg-ratio checks for cbBTC, WBTC and LBTC, plus circuit-breaker functionality that can suspend new borrowing when collateral prices diverge from reference values. Supply caps and rate limits are additional mitigants, but their present utilization and effectiveness are Not verifiable as of October 10, 2026.
• Depositor loss path: The described framework places losses first against Sky surplus/reserves and other backstops, with residual losses potentially socialized across USDS holders; this is a protocol-design assertion, not an independently verified solvency result.
• Chain concentration: A recent analytics source reported approximately $7.96 billion on Ethereum, $325.3 million on Base, and $10.0 million on Arbitrum; Avalanche and Robinhood Chain were reported at zero or unavailable. These figures are aggregator data, not raw on-chain verification, and are stale relative to the required latest-block standard. Key limitation: Current BTC collateral balances, Spark Savings exposure by chain, USDS reserves, and post-liquidation residuals are Not verifiable as of October 10, 2026. The stress appears low-to-moderate for direct BTC exposure, but tail risk remains through USDS/Sky balance-sheet contagion and liquidity withdrawals.

Assessment: Not verifiable as of 2026-10-03. A quantified loss from a 20% depeg of Spark Savings’ largest collateral asset cannot be calculated from the available public evidence because the current collateral composition, largest-asset concentration, and chain-level exposure are not verifiable without on-chain data. The result therefore remains unchanged from the 2026-09-03 finding. The relevant public risk disclosures indicate that multiple Savings products—including spUSDC, spUSDT, and spPYUSD—are backed by USDS and share a common collateral portfolio, so a depeg of a major underlying collateral asset could affect multiple vaults rather than only the entry asset.

Credora’s latest publicly indexed assessment reports differentiated stressed-loss probabilities across vaults, but it does not provide the exposure data needed to translate a 20% collateral shock into a protocol loss. Stress transmission and mitigants:

• Spark may place Savings vaults into recovery mode, including temporarily halting or coordinating withdrawals to reduce bank-run and liquidity risks.
• Loss absorption is described as sequential: Prime-level/internal and external junior risk capital, senior or surplus reserves, and other backstop facilities are used before residual losses reach depositors.
• If those layers are exhausted, Sky may mint SKY as a token backstop; only after backstops and other risk capital are exhausted would residual losses be socialized among USDS holders, including Savings vaults.
• These mechanisms reduce the expected depositor impact but do not establish that a 20% depeg would be fully absorbed, nor do the sources disclose presently available buffer capacity sufficient to prove that outcome. Contradiction / evidence gap: Spark documentation describes “several hundred million dollars” of protection, while the available evidence does not independently verify current buffer size, collateral weights, or the portion attributable to each listed chain. The documented protection claim is therefore not a quantified stress result and should not be treated as proof of solvency under this scenario. On-chain collateral composition, largest-collateral exposure, chain percentages, and resulting loss: Not verifiable as of 2026-10-03.

Stress scenario: top external counterparty becomes insolvent *(verified against sources available on October 4, 2026; Dune unavailable).* Expected loss path 1. The Spark Liquidity Layer (TAKER_ROLE) has drawn assets from the relevant ERC-4626 vault into the counterparty/strategy using take(). Insolvency prevents the expected assets and yield from being returned.

2. On-chain, the vault’s totalAssets() remains share-implied while asset.balanceOf(vault) falls short; assetsOutstanding() rises. Withdrawals/redemptions above idle liquidity revert with SparkVault/insufficient-liquidity.

Liquidity Intents may source replacement liquidity, but do not eliminate the economic loss. 3. The documented loss waterfall then applies: Prime internal junior risk capital, Prime-external junior capital, the Sky Surplus Buffer, aggregate/Genesis backstop capital where applicable, and finally SKY-token recapitalization.

Only after those layers are exhausted is residual loss socialized across USDS exposure, including USDS-backed Spark Savings vaults. Who absorbs it / compensation

• First absorber: the relevant Prime/allocated risk capital—not depositors directly.
• Next absorbers: Sky surplus and ecosystem backstop capital.
• Tail case: SKY issuance/recapitalization; any remaining deficit can reduce the value of USDS exposure. The newer framework additionally describes possible SKY distributions to affected holders, but this is not specified as an automatic depositor-insurance payment.
• Guaranteed compensation or fixed insurance payout: Not verifiable as of October 4, 2026. The ETH vault is described as backed by Spark’s balance sheet and insurance, but coverage terms and claim mechanics are not provided. Smart-contract impact The vault does not automatically haircut shares when a counterparty defaults. Its share-implied liability can remain above immediately available assets; withdrawals are liquidity-constrained until capital is returned or injected. Governance can set the deposit cap to zero or upgrade the UUPS implementation, but V2 has no native pause/freeze/recovery flag. > Contradiction / caution: Spark’s current product docs say srUSDS is still planned, while the newer security-framework article describes a Genesis Capital Backstop and SKY distributions. Treat the exact intermediate waterfall and compensation mechanics as not fully contractually verified. Chain exposure: The loss logic is common across deployed vaults; chain-level exposure percentages for Arbitrum, Avalanche, Base, Ethereum, and Robinhood Chain are Not verifiable as of October 4, 2026 without Dune/on-chain data.

As of October 10, 2026, I found no credible independent evidence that Spark Savings or its governing Sky/Spark DAO, Phoenix Labs, or identified owners committed fraud, misappropriated user funds, or operated a rug pull. This conclusion is limited: the available search did not establish a regulator finding, court judgment, or independent forensic investigation proving the negative. On-chain verification is unavailable in this run; therefore, ownership, treasury movements, and fraud-related transaction patterns are Not verifiable as of October 10, 2026.

The strongest available evidence is an independent risk review reporting no documented fraud, manipulation, or rug-pull architecture, while identifying governance centralization and upstream Sky/MakerDAO dependency as structural risks rather than fraud allegations. A separate analysis likewise distinguishes Spark Protocol from unrelated projects such as Sparkster and CoinSpark that faced enforcement or scam allegations. The ChainSecurity assessment found a withdrawal-request spam/economic-barrier issue in Spark Savings Intents; this is a smart-contract design weakness, not evidence of committed fraud.

The relevant adverse scenario is governance abuse or upstream failure, not established fraud: Sky governance reportedly retains significant control over Spark’s parameters and assets, creating concentration and conflict-of-interest risk. That risk should not be relabeled as proof that DAO members or owners committed fraud. Finding: No substantiated fraud finding identified. Fraud allegations, legal proceedings, owner enrichment, and DAO-controlled fund flows are Not verifiable as of October 10, 2026.

Protocol-affiliated statements or documentation were not used as proof of innocence. Search results also contained unrelated same-name entities; those were excluded under the name-collision rule.

Stress assessment — primary yield source negative over 30 days: Spark Savings’ primary source for USD-denominated savings exposure is Sky’s Savings Rate (SSR; DSR for legacy DAI exposure). The rate is funded by Sky borrowing fees and set by Sky Governance, rather than generated by an independent Spark market-neutral strategy. [https://docs.spark.finance/products/spark-savings] A sustained negative 30-day source yield is not a documented operating case. Available evidence states that governance may reduce the SSR to zero or another non-negative value, but does not document a negative savings-rate mechanism. [https://eco.com/support/en/articles/15254003-sky-savings-rate-deep-dive-2026-ssr-susds-usds-mechanics] Stress result: the realistic documented failure mode is yield collapse to zero or near-zero, not negative carry mechanically charged to depositors.

If Sky’s underlying economics became loss-making, the residual risks would instead be reduced distributions, governance repricing, impaired reserves, or principal/solvency losses; the reviewed materials do not establish a contractual floor protecting Spark users from those broader risks. [https://eco.com/support/en/articles/14798655-susds-explained-sky-s-savings-rate-token] Risk conclusion: classify this scenario as high impact / low observability. A 30-day negative primary-source yield cannot be verified as an implemented SSR behavior. The appropriate modeled shock is 0% gross yield for 30 days, plus a separate tail scenario for loss of principal or sUSDS/USDS impairment.

The reviewed sources do not provide chain-specific exposure, vault-level loss allocation, or a verified 30-day negative-yield event. Not verifiable as of October 3, 2026.

Assessment (as of October 13, 2026): Spark Savings is not an independent DAO. It operates as a Spark/Sky SubDAO product: Spark/Snapshot proposals authorize changes, but Sky Executive Votes and spells execute them. Token-holder governance is therefore substantive at the Sky/Spark level, but not autonomous at the individual Spark Savings product level. Control surface: Ethereum vaults identify the Spark SubDAO Proxy as DEFAULT_ADMIN; the ALM Proxy is the taker/custodian, while an ALM Proxy Freezable or operations safe controls operational parameters.

Arbitrum deployments use a Spark Executor as admin and an ALM Proxy/Freezable for operations. The ALM system’s documented admin can upgrade controllers, set rate limits, grant/revoke roles, and recover funds; relayers can execute only constrained, rate-limited operations. Proposal process: eligible Spark proposers submit Forum proposals; risk/operational review precedes a three-day Snapshot vote. Successful proposals are encoded in the Sky Atlas/Spark spell process and executed through Sky governance. Timelock: Sky’s GSM Pause Delay is currently 48 hours, replacing the previously recorded 24-hour value.

This is a governance execution delay, not proof that every vault parameter has an independent vault timelock. Multisig/signers/concentration: signer identities, thresholds, independence, voting concentration, and top holders require Dune/on-chain verification. Not verifiable as of October 13, 2026. Frontend: the interface is an open-source repository with IPFS deployments; effective deployer/admin control is not independently verified. Company entity, jurisdiction, registration number, directors, and applicable ToS are Not verifiable as of October 13, 2026. Risk conclusion: DAO governance is real at Sky/Spark governance level but product-level control is centralized through SubDAO/admin, ALM, and operational roles. Administrative governance can ultimately recover/reconfigure funds; operational takers can move assets within configured permissions and limits.

Assessment (as of October 4, 2026; off-chain review only). On-chain legal-entity attribution is Not verifiable as of October 4, 2026 because Dune was unavailable. Entity / structure. The Savings website/front end identifies Spark.fi Inc. as its owner/operator, incorporated in Panama with a Panama City registered office. Its privacy policy names Spark.fi Inc. as the data controller. Separately, the MiCA materials identify Spark Foundation (Cayman Islands) for SPK admission to trading and **SPK Company Ltd.

(British Virgin Islands) for the SPK offer. These token entities should not automatically be treated as the Savings product’s operating entity. Terms, restrictions and classification.** The Terms describe Spark as decentralized, open-source smart-contract software and state that the site operator does not custody assets, control transactions, or act as an intermediary. Users must be legally eligible, comply with applicable law, and not use the site from jurisdictions where the activity is illegal, prohibited, or requires licensing; sanctions circumvention is prohibited.

The Terms expressly state that neither the site operator nor the technology is registered, licensed, or actively supervised by a financial regulator, and that the operator is not a broker, dealer, adviser, transfer agent, or other intermediary. Governing law and court jurisdiction are Cayman Islands, despite the front-end operator being Panamanian. KYC/AML and data protection. The privacy policy permits collection of identity, address, nationality, date of birth, wallet and other personal data, including for KYC, AML/CTF, PEP and sanctions screening. It does not establish that every Savings user must complete KYC; the practical trigger and scope are Not verifiable as of October 4, 2026.

Panama Law 81/2019 and Executive Decree 285/2021 are identified as the principal data-protection framework; cross-border transfers and third-party analytics are disclosed. Warnings / enforcement / litigation. The MiCA white papers are expressly stated not to be approved by an EU competent authority. No protocol-specific regulator enforcement, sanctions designation, or court case was identified in the reviewed sources; unrelated entities with “Spark” in their names were excluded. Formal product classification as a deposit, security, investment product, or licensed financial service is Not verifiable as of October 4, 2026. Legal structure vs. actual risk: Panama front-end entity, Cayman contractual forum, and separate Cayman/BVI token entities create entity- and recourse-fragmentation risk.

Decentralization language does not eliminate possible regulatory, consumer-protection, AML, securities, or insolvency exposure at the operator, contributor, governance, intermediary, or underlying-strategy levels. Structured fields:

• active_enforcement: false
• sanctioned: false
• entity: Spark.fi Inc. (front-end/site operator); separate token entities: Spark Foundation and SPK Company Ltd.
• jurisdiction: Panama for Spark.fi Inc.; Cayman Islands for Spark Foundation; British Virgin Islands for SPK Company Ltd.

No exact GLEIF LEI record for 'Spark.fi Inc', 'separate token entities: Spark Foundation', 'SPK Company Ltd', 'Spark Savings'. OFAC SDN screening of 'Spark.fi Inc', 'separate token entities: Spark Foundation', 'SPK Company Ltd', 'Spark Savings': no match. SEC litigation and administrative release feeds: no mention.

Spark Savings does not issue its own stablecoin; it is a savings wrapper that accepts external assets, and the docs explicitly say USDS is not issued or deployed by Spark. The stablecoins directly used in Spark Savings are USDC and USDS (plus USDT, PYUSD, and ETH in newer vaults), but a protocol-specific historical depeg count for Spark Savings itself is not verifiable as of 2026-10-05. The only clearly documented depeg risk in the retrieved material is USDS depeg risk in general, and no on-chain verification is available in this run; therefore depeg_count, last_depeg_date, and max_depeg_pct are not verifiable as of 2026-10-05.

Spark Savings has material exposure to allocator/counterparty, liquidity, issuer, governance, and integration risks. DeFiLlama’s latest indexed data reports approximately $1.321 billion TVL across eight chains, with Ethereum representing 74.7%; the user-specified five chains omit OP Mainnet, Unichain, and X Layer. On-chain contract-state verification is Not verifiable as of October 5, 2026.

Spark Savings’ top 5 strengths are: 1) institutional-scale liquidity and fast withdrawals, with 24/7 redemptions, large-deposit support, and asynchronous withdrawal intents for bigger requests; 2) deep multi-protocol liquidity access through the Spark Liquidity Layer, which coordinates liquidity across Spark Savings, SparkLend, the Sky PSM, and other approved venues; 3) stablecoin-focused yield, making it straightforward to earn on stablecoins without active trading or leverage; 4) multi-chain availability, with support across chains such as Ethereum, Base, Arbitrum, Avalanche, and Robinhood Chain; and 5) a risk-managed design emphasizing backstop facilities, liquidity buffers, and non-custodial savings vaults. The newest available third-party and protocol-adjacent sources do not materially change these strengths versus the prior check, though the “security/transparency” point is still best treated as a design claim rather than independently verified on-chain evidence in this run. Not verifiable as of 2026-10-03 for on-chain confirmation.