Symbiotic

Tier 2

Security Analysis Report

TVL

$306.15M

Sections

7

report sections

Citations

528

evidence references

Version

v3

10/29/2026

5.4

Moderate Risk

Safety Score · out of 10

📋

Executive Summary

5.4/10

Executive Summary: Symbiotic

Symbiotic is a large Ethereum-only "collateral markets" protocol with a named team, many audits and no recorded hack. Its main weaknesses are structural: the audits are hard to link to the code actually deployed, a small multisig with no delay controls the core factories, and most deposits sit in a few BTC wrappers and ETH staking tokens with no loss-cover fund. The evidence is moderately strong, but vault-by-vault exposure is not published. No hack is recorded, so these are exposures, not realised losses.

Safety Rating

Overall Score: 5.4/10 Dimension scores: Code security 4.3 · Economic security 5.3 · Team transparency 7.5 · Governance 2.8 · Track record 8.0

Scores by dimension, points out of 10 (points)
Code security
4.3 points
Economic security
5.3 points
Team transparency
7.5 points
Governance
2.8 points
Track record
8.0 points

Scores by Dimension

DimensionKey Finding
Code SecurityMany audits exist, but audited revisions are rarely tied to deployed contracts.
Economic SecurityDepositors rely on curator choices and third-party strategy layers, and no insurance or reserve fund is identified.
Team TransparencyThree leaders are named with professional histories, but the legal entity and treasury reporting are not disclosed.
GovernanceA 2-of-3 Safe with no timelock controls the Core factories, and no voting process was found.
Track RecordOperating since 2024-06-11 with no incident in the hacks dataset; the record is absence of evidence, and the V2 core is much newer.

What This Protocol Does

Symbiotic lets users deposit assets into curator-run vaults whose capital also backs obligations such as loan guarantees or insurance cover. DefiLlama shows about $316.6 million of TVL (total value locked), all on Ethereum. Depositors take on the risks of their curator's strategy, including slashing (a penalty that takes part of the collateral).

The product set covers Core vaults (V2, with a legacy layer still on-chain), adapters, the Liquid Lane redemption product, Rewards and the Relay signing network. Cap (credit) and Nexus Mutual (insurance) are applications built on top.

Key Strengths

  • Broad audit program, with uneven depth by product. There are 35 audit reports from about 12 firms, a Cantina competition, Certora formal verification of the earlier Core version and a listed Cantina bounty of $500,000. The bounty's live status is not confirmed. Twenty of the 35 reports sit on hosts not verified as auditor-controlled; this lowers confidence, not the score.
  • Overlapping review of Liquid Lane. Bailsec and OpenZeppelin both report the same Liquid Lane issues. The OpenZeppelin report sits on a host not verified as auditor-controlled, so the overlap shows attributed coverage, not independent confirmation. This evidence covers Liquid Lane only. Certora formal verification and the Cantina competition cover the earlier Core, not Core V2. The Relay's critical items are open (see Key Risks).
  • Core V2 review fixes (Core V2 only). Statemind's Core V2 review marks all 5 medium findings fixed. It is an affiliate review and does not show that the fixes are deployed.
  • Named, experienced team. The CEO and CTO co-founded Statemind, which is also an auditor of Symbiotic.
  • Split registry control and zero protocol fee. A separate 3-of-5 Safe owns the adapter registry. Global fees read 0 on-chain at block 26296259. The caveat: the 2-of-3 Safe can change fees and limits with no delay.

Key Risks

  • 1. Core V2 vault withdrawals can freeze or leak value through adapters and oracles (Core V2 vaults, Liquid Lane vaults, Morpho/Aave/Euler-adapter vaults). Magnitude: a freeze, not a certain loss, with no stated maximum duration; the share of TVL affected is not published. Likelihood: medium for a temporary freeze in at least one vault, low for a profitable exploit. One frozen redemption account can block deallocation and all vault withdrawals (acknowledged, not fixed). These findings concern audited or public repository code; their presence in deployed vaults is not established. Donation attacks stay open on adapters other than Morpho (partially fixed). No maximum delay is stated for Core V2 vaults.
  • 2. Concentrated collateral in BTC wrappers and ETH staking tokens, with no loss-cover fund (uniBTC, LBTC, wstETH, rETH, weETH vault deposits). Magnitude: uniBTC and LBTC together are about USD 143M, roughly 45% of TVL. The ETH staking tokens (wstETH, rETH, weETH, osETH and sfrxETH) add about 42% by share. A failure would be severe because the exposure is correlated. Likelihood: low, since it needs an external failure at an issuer, custodian or bridge.
  • 3. A 2-of-3 Safe with zero delay and unnamed signers controls factory versions, while the deployed VaultFactory differs from its audited revision (new vaults and vaults whose owners migrate; VaultFactory has 367 entities). Magnitude: not quantified; the harm would be a malicious version reaching new vaults or vaults whose owners migrate, because a migration likely needs both the vault or adapter owner and a whitelisted version. The review found no direct fund-seizing path in the fee and limit functions. Likelihood: low, because two signers are needed and no abuse is recorded. On-chain at block 26296259 (2026-10-29) the whitelist call has zero delay. The deployed VaultFactory differs from the 2024-09 OtterSec-audited version in 4 files, judged material. The 2024 Core audit says a whitelisted implementation cannot be removed; whether this holds for the current VaultFactory is not confirmed.
  • Further risk: open critical findings in the Relay (separate from Core vault deposits). A 2025 AI-generated Statemind report's missing BLS subgroup check (a signature-input check) is still present in the latest public relay-contracts code; deployment was not compared. The other three critical items were not located or not checked. Likelihood is low to medium; the harm is severe for networks relying on Relay, minor for Core vault depositors.
  • Further risk: audits are weakly linked to deployed code. Only 1 of 25 deployed component rows is shown identical to an audited version; the others differ, have no identified audited revision, or have an undetermined scope. For 77.8% of compared lines the audited version was not identified, so these lines are "not compared", not shown to be changed.

Critical Findings

  • No hack or loss is recorded.
  • Four critical findings from the 2025 AI-generated Statemind Middleware report are open; one (missing BLS subgroup checks) and three high-severity items are still present in the latest public Relay code, with deployment not compared.
  • Across all audits, 108 issue groups rated medium or above are not marked fixed (4 critical, 25 high, 79 medium).
  • Documented versus on-chain: LlamaRisk (August 2024) says the core contracts are non-upgradeable. The Core V2 and Liquid Lane audits describe a migratable pattern, and on-chain at block 26296259 (2026-10-29) a 2-of-3 Safe owns the factories, so current immutability is not established.

Recommendation

Assessment: Symbiotic suits risk-tolerant users who can judge a specific curator, vault and adapter, and who accept possible withdrawal freezes and issuer risk in BTC wrappers and staking tokens. Check the audit coverage of the exact vault before depositing, because the evidence does not show that every deployed contract matches an audited revision. This view rests on the reported strengths in team disclosure and audit volume and the weaknesses in factory-level delay and audit-to-deployment traceability. Users with low risk tolerance should wait for a factory-level delay, named signers and clearer audit-to-deployment links. The factory powers have no delay and no voting venue gives a community check. The no-hack record does not show that Core V2, launched in July 2026, has been stress-tested.

Report Sections

  • Protocol Overview
  • Code Security
  • Economic Security
  • Social & Team
  • Governance
  • Cross-Risk Analysis

45 citations · click markers to view evidence

🏗

Protocol Overview

Protocol Overview: Symbiotic

Symbiotic is a "collateral markets" platform on Ethereum: people deposit assets into curator-run vaults, and applications such as credit, insurance and tokenized-asset redemption services use that capital as code-enforced backing for their obligations. The DeFiLlama record shows about $316.6 million of TVL, all on Ethereum, in the category Collateral Markets. Depositors earn yield from whatever strategy their vault's curator chooses, and they also take that strategy's risks, including slashing (a penalty that takes part of the collateral). Admin control of the core factories sits with a 2-of-3 Safe multisig with no measured timelock, so users trust that small group as well as the code. The evidence is strongest for Core V2 mechanics and weakest on exact withdrawal rules and on per-vault yields.

What It Does

Symbiotic lets users deposit crypto assets into vaults whose capital also backs financial obligations, for example loan guarantees or insurance cover. The same capital keeps earning while it is committed. It originally launched as a restaking marketplace (restaking means reusing staked assets to secure several networks). It now describes itself as a collateral-markets platform, and the docs say Core V2 launched in July 2026.

  • Category: collateral markets / restaking infrastructure. DeFiLlama files it under Collateral Markets.
  • Product set in this record: Core vaults (V2, with a legacy DefaultCollateral layer still on-chain), adapters, the Liquid Lane redemption product, Rewards, and the Relay (a stake-backed signing/attestation network). Credit (Cap) and insurance (Nexus Mutual) are applications built on top.
  • Participants:
    • Allocators deposit capital.
    • Curators design vault strategy.
    • Applications use the capital.
    • Operators do work that the collateral backs.
  • Use cases: credit guarantees, insurance underwriting, instant redemptions for tokenized real-world assets (RWAs), and stake-backed attestations for bridges and appchains.

How It Works

Capital flows from allocators into a vault, through the vault's Universal Delegator into adapters, and from there into applications or external DeFi protocols.

  1. A curator deploys a Vault V2 through the VaultFactory. The script also creates the share token and a linked Universal Delegator. Vault creation is permissionless.
  2. An allocator deposits the vault's single asset and receives ERC-20 vault shares. Deposit limits or a depositor whitelist may apply.
  3. The curator adds adapters to the vault's UniversalDelegator, then allocates assets across them.
  4. Adapters put the capital to work. They can be native Symbiotic applications such as Cap, or external strategies such as Morpho and Aave, or Liquid Lane redemptions.
  5. While capital backs an obligation, the application can slash it if a defined loss event occurs.
  6. As the strategy earns, vault shares rise in value, and the allocator withdraws through the vault. The Bailsec Vault V2 review describes two exit paths. A user can redeem instantly only after the queue priority is satisfied. Otherwise the user moves shares into the WithdrawalQueue, receives a withdrawal NFT, and later claims assets as the queue is filled. Instant liquidity is limited to the vault's idle assets plus adapter assets that can be deallocated at once. This describes the code at the reviewed commit. Exact withdrawal timing and equivalence with the deployed code are not documented (see Data Completeness).
Vault V2 capital path from allocator to adaptersOpen full view →
ACTORAllocatorACTORCuratorCONTRACTVault V2 (ERC-20shares)CONTRACTUniversal DelegatorCONTRACTAdaptersdeposits asset,gets shares2adds adapters,allocates3uses to allocateassets3routes capital4
  • (actor)
  • (actor)
  • (contract)
  • (contract)
  • (contract)
  • step 2flow
  • step 3control
  • step 3flow
  • step 4flow

Core mechanism. Vaults act as the coordination layer: they hold allocator capital and enforce the curator's rules. Collateral is committed for defined periods, and it cannot leave while the obligation is active.

Slashing. Slashing is how a network penalizes operators for provable misbehavior. The Slasher module checks a penalty request and does not detect faults itself. A vault is deployed with one of three slashing modes, and the choice is fixed afterwards.

  • No slashing.
  • Instant Slasher.
  • VetoSlasher, which adds a veto window in which resolvers can block the slash.

A slash is capped by the stake guarantee captured earlier. Slashed collateral goes to a Burner, which can burn, redistribute or route it. These slashing pages sit in the Universal Staking (earlier-generation) documentation, and the docs read do not show how Core V2's AppAdapter maps onto them.

Slashing path for a vaultOpen full view →
ACTORNetwork or applicationmiddlewareCONTRACTSlasher or VetoSlasherROLEResolver (veto only)CONTRACTBurnerrequests cappedslashcan veto in windowhands slashedcollateral
  • (actor)
  • (contract)
  • (role)
  • (contract)
  • risk
  • control
  • flow

Liquid Lane (redemption product). Market makers use curator-managed vault liquidity to buy back tokenized assets that are otherwise slow to redeem. Asset-manager steps in the docs:

  • A swap request arrives via a frontend or RFQ (request-for-quote) system.
  • The market maker's contract pulls USDC from the vault.
  • USDC is released to the asset manager atomically in the same transaction.

Relay. Operators sign messages, and the signatures are combined with BLS aggregation, backed by bonded stake. The Settlement contract stores validator-set headers by network, subnetwork and epoch.

Key contracts (Ethereum mainnet). Counts come from on-chain reads at block 26296259 (2026-10-29 05:19 UTC).

ContractRoleMeasured on-chain
VaultFactoryCreates and migrates vaultsversion 3, 367 entities
UniversalDelegatorFactoryCreates delegators32 entities
WithdrawalQueueFactoryCreates withdrawal queues32 entities
MorphoVaultV2AdapterFactoryMorpho adapters19 entities
LiquidLaneAdapterFactoryLiquid Lane adapters17 entities
NetworkRegistryRegistered networks82 entities

Other factories exist for Aave V3, Euler and 3F adapters, and for per-issuer Liquid Lane account contracts, such as JAAAAccountFactory. The source-provided address list is in the official addresses page.

Yield Sources

Vault yield depends on the curator's strategy and comes from several sources. Based on available data, no per-vault APY history is provided.

  • Base lending yield. Adapters to Morpho, Aave and Euler earn lending interest.
  • Liquid Lane spread. Market makers pay for instant redemptions. The docs say this is driven by redemption demand (early exits and liquidations), and the average vault lockup is 14 days.
  • Application payments. Credit and insurance applications pay for the guarantee. In the Cap case, allocators receive fees for providing the guarantee and taking credit risk. Nexus Mutual underwriters earn premiums.
  • Network rewards. These may come from fees, token emissions or client payments.
  • Points. Core V2 points accrue at the rate below, and the docs do not describe them as a payout.

Fees that reduce depositor yield. Vault V2 curators can charge a management fee, capped at about 5% a year, and a performance fee, capped at 20%. Both are paid by minting new vault shares, which dilutes depositors. The protocol fee registry shows a zero global management fee and a zero global performance fee on-chain at block 26296259.

Sustainability. Assessment: the Liquid Lane spread, lending interest and application fees are earned from real usage, not only from emissions. Those sources depend on redemption demand, lending rates and application volume, so they can fall. Points are a separate incentive. No historical yield range exists in the sources.

Dependencies & Integrations

Symbiotic depends on the external protocols its curators choose, the assets that vaults accept, and the applications that can slash capital.

  • Oracles: the Relay contracts include a ChainlinkPriceFeed helper for pricing voting power. The price source for Liquid Lane redemptions is not identified in the sources.
  • DeFi protocols: Morpho, Aave and Euler adapters.
  • Tokenized-asset issuers (Liquid Lane):
    • Midas
    • Centrifuge
    • OpenEden
    • Hastra / Figure
    • Pareto
  • Applications:
    • Cap, a covered-credit platform.
    • Nexus Mutual, a decentralized insurance protocol.
    • Hyperlane.
  • Collateral tokens: the largest TVL shares are uniBTC, wstETH and LBTC (see the chart under TVL & Adoption). Each token carries its own issuer, contract and bridge risk, which Symbiotic vaults inherit. Assessment: the heavy share in BTC wrappers and liquid staking tokens makes Symbiotic exposed to those issuers.
  • Chains: Ethereum only in the DeFiLlama record. No bridge dependency is documented for Core. The Relay is designed to verify attestations on other chains.

TVL & Adoption

DeFiLlama captured $316.6 million of TVL for Symbiotic, all of it on Ethereum. The snapshot was collected on 2026-10-29. The same snapshot shows a 7-day change of about -4.5%. No older durable TVL series is in the sources.

The dependency graph attributes TVL to these tokens (composition measured 2026-10-29). The largest shares are uniBTC at 30.38% , wstETH at 25.79% , LBTC at 14.7% , rETH at 6.86% , weETH at 6.08% and SPK at 4.18% .

TVL composition by token, share of total percent, 2026-10-29 (percent)
uniBTC
30.38 percent
wstETH
25.79 percent
LBTC
14.7 percent
rETH
6.86 percent
weETH
6.08 percent
SPK
4.18 percent

Notable integrations: Cap (docs say Symbiotic powers more than $250m of collateral for borrowers) , Nexus Mutual, and Liquid Lane issuers including Midas and Centrifuge. The website banner cites $1.6B AUM on Centrifuge funds that gain instant liquidity (the banner's figure is for the funds, not Symbiotic TVL).

Technical Architecture

The factory and fee-registry contracts are owned by a 2-of-3 Safe multisig, and no timelock delay was measured. Assessment: users must trust that signer group not to whitelist harmful vault versions or change fees, because there is no on-chain waiting period.

Who controls what (Ethereum, block 26296259, 2026-10-29 05:19 UTC).

  • Factories. The VaultFactory is owned by a Gnosis Safe with 3 owners and a threshold of 2. The same Safe also owns the delegator, adapter and account factories, and the protocol fee registry.
  • Vault versions. The Safe can whitelist vault versions and blacklist version numbers, each with a measured fastest delay of 0 seconds (no timelock). A vault owner can migrate to versions Symbiotic whitelisted. The parser could not resolve who may call migrate (delay unknown).
  • Fees. The same Safe can set the global protocol fee.
  • Registries. AdapterRegistry and the Liquid Lane AccountRegistry are owned by a separate Safe with 5 owners and a threshold of 3. The AdapterRegistry owner is that Safe's address. That Safe is not in the source-provided contract list.
  • Signers. A web check on 2026-10-29 read the owner lists of both Safes from the Safe client gateway (source reliability: unverified publisher). The 2-of-3 Safe lists these owners: One of them, 0x62f8…6C4c78, carries an Etherscan label "Symbiotic: Deployer". That label marks a Symbiotic-associated account, not a named person. The other two owners have no public name tag. The 3-of-5 Safe lists five owners, including this address. None of the five carries a public tag. No signer of either Safe is attributed to a named person or entity in the Symbiotic docs or homepage, and ENS name lookups failed, so the people behind the keys remain unknown.
  • Per vault. Curators control the strategy, fees, deposit limits and adapters. The default script gives the vault owner all roles.

Upgradability. No standard proxy implementation or admin slot was found for the factories at this block, which is not proof they cannot be changed. An August 2024 LlamaRisk report, written about the earlier V1 core, said core contracts were non-upgradeable. The V2 vault-migration path above is the main change mechanism.

Emergency mechanisms. The sources reviewed show no global pause or shutdown. A Cantina bug bounty with a $500,000 maximum covers Core and Rewards.

Who controls the Ethereum core factories and registriesOpen full view →
GOVERNANCESafe 2-of-3 (factoryowner)GOVERNANCESafe 3-of-5 (registryowner)CONTRACTVaultFactoryCONTRACTProtocolFeeRegistryCONTRACTAdapterRegistryROLEVault owner / curatorwhitelists vaultversionssets globalprotocol feeownsmigrates towhitelisted…
  • (governance)
  • (governance)
  • (contract)
  • (contract)
  • (contract)
  • (role)
  • control
  • control
  • control
  • control

Key Risks (Protocol-Level)

User funds sit in curator-managed vaults and in whatever adapters, applications or external protocols the curator picks, so vault choice drives risk.

  • Strategy and adapter risk. Liquidity adapters add operational and financial risk. Liquid Lane adds issuer and redemption risk for tokenized assets.
  • Slashing risk. Capital backing credit or insurance can be slashed on a default or valid claim. The application decides when a loss occurred.
  • Admin-key risk. A 2-of-3 Safe with no measured timelock controls factory whitelisting and fees (see Technical Architecture). Assessment: the likelihood of abuse is low to moderate and the impact is broad across 367 vault entities. The basis is a two-signer quorum, a zero measured delay and signers with no public attribution to named people.
  • Relay design risk (separate component from Core vaults). A Sherlock contest finding says an operator could manipulate the validator set after unstaking and avoid slashing. The claim records that the team acknowledged it and would not fix it at that time. The same contest also reports a proof-forgery issue; its fix status is not shown here. These findings concern the Relay, not Vault V2.
  • Complexity. There are many factories, adapter types and per-issuer account contracts. Some Core V2 components have no attributed current address in the contract list, so their on-chain controls were not checked.
  • Audit depth. Two audits cover Vault V2 directly. Statemind reviewed Core V2 from 28 May to 26 June 2026; the scope includes the Vault V2 system, Universal Delegator and adapters. It found 41 issues: 5 medium, all fixed, and 36 informational, of which 14 were fixed and 22 acknowledged. The medium findings concern AppAdapter withdrawal debt and slashing. One lets an attacker brick slashing through an unbounded claimable backlog.
    • Bailsec, June 2026. The "Vaults V2" report found 60 issues and lists its repository path as the src/contracts folder of the Symbiotic core mirror. The audit index does not state a file list, so the exact files are not confirmed. Of the 3 high findings, 2 were resolved and 1 partially resolved. Of the 8 medium findings, 1 was resolved and 7 were acknowledged with no change made. Of the 29 low, 19 informational and 1 governance findings, 28, 16 and 1 were acknowledged with no change. These statuses are as of the audit, which applies only to the specific commit hash, version or deployment reviewed. Whether the partly resolved high finding or the acknowledged mediums are present in deployed vaults was not compared. The Bailsec report states that centralization and admin-key risks are outside its scope.
    • Core V2 Extension. The official audit page also lists a Bailsec review titled Core V2 Extension (June 2026). That page gives titles only, so its scope and findings are not summarised here.
    • Liquid Lane audits (separate product from Core vaults). The audit host of each report below is not verified as auditor-controlled, so the findings carry lower source confidence. Each audit counts only for the Liquid Lane components in its scope, not for Vault V2.
      • OpenZeppelin, June 2026 (Core v2 Adapters). The review covered the LiquidLaneAdapter, its redemption-account hierarchy, the CoW conversion module, the Midas oracle and the shared migratable base. It reported 28 issues, of which 11 were resolved and 2 partially resolved. None was critical or high. Of 7 medium issues, 1 was resolved and 2 partially resolved. The auditor names the most consequential issue as availability: a single token whose oracle returns zero or reverts can brick vault accounting.
      • Bailsec, June 2026 (Liquid Lane Adapters). It reported 46 findings: 4 resolved, 2 partially resolved and 40 acknowledged with no change. The audit index could not confirm the file scope of this review.
      • Bailsec, July 2026 (Async/Account scopes). Scope 1 covers MidasAccount, with 10 findings: 3 resolved, 1 partially resolved and 6 acknowledged. Scope 3 covers AsyncRedeemAccount, with 14 findings: 3 resolved, 1 partially resolved and 10 acknowledged. Scope 4 covers FigureAccount and OpenEdenAccount, with 16 findings, all acknowledged with no change.
      • Coverage gap. Whether the deployed Liquid Lane contracts match the reviewed commits was not compared.
    • Not covering Core V2. Audits of Relay, Rewards and earlier Core versions do not cover Core V2. Whether the deployed code matches the audited commits was not compared.

Data Completeness

  • Available: official docs for Core V2, curators, fees, slashing, Liquid Lane and use cases; the DeFiLlama snapshot (2026-10-29); on-chain control reads for Ethereum at block 26296259; claims from audit digests.
  • Missing or limited:
    • Withdrawals. The Bailsec review describes the instant and queued exit paths, but the time a queued withdrawal takes to fill is not documented. Whether the deployed vaults match the reviewed code was not compared.
    • Yields. There are no per-vault APY figures and no historical yield series.
    • TVL history. No dated TVL history exists beyond the single snapshot.
    • Signers. The owner addresses of both Safes were read, and only one carries a public label (a Symbiotic deployer account). Named people, signing processes and ENS names are not confirmed.
    • Contract coverage. Many on-chain reads were cut short by time or address limits, and many Core V2 contracts lack attributed addresses.
    • Adoption claims. Counts of stakers, operators and vaults on the protocol's listing are not backed by a citable source here.
    • Articles. No third-party research articles were available. The only research report, LlamaRisk (August 2024), is old and describes the earlier V1 core.
    • Oracle provider. The price source for Liquid Lane redemptions is not identified.
    • Other chains. Only Ethereum is active in the DeFiLlama record. No other chain deployments were assessed.

134 citations · click markers to view evidence

🔒

Code Security

4.3/10

Code Security Analysis: Symbiotic

Symbiotic has a broad external audit program: 35 audit reports from about 12 firms, plus a Cantina competition, formal verification by Certora on the earlier Core version and a Cantina bug bounty listed at $500,000. The weak point is traceability. For most deployed contracts the audited revision could not be identified, and 77.8% of compared lines have no identified audited version. Four critical findings from a 2025 AI-generated Statemind report on the Symbiotic Middleware codebase at commit 5c801a0 are open. The code check matched that report to the relay-contracts repository and found one critical flaw (missing BLS subgroup checks) and three high-severity items still present in its latest code; the other three critical items were not located or not checked. Users should check audit coverage for the specific vault and adapter they use, because the evidence does not show that every deployed contract matches an audited revision.

Data Completeness

Findings: 587 findings from the 35 audit reports, grouped into 520 issues. Of these, 108 groups rated medium or above are not marked fixed (4 critical, 25 high, 79 medium). Sources read: the list of audit findings with their status, the audit reports' scope information, the deployed-code comparison data, the results of the code checks against the latest public repositories, the summaries of the Statemind Middleware and Certora reports, the docs audit page and the audit sections quoted below. Audit text beyond those sections reached this report through the findings list, not by a full read of every report.

  • Code checks were run for 24 of the 135 findings selected for checking; one check result (In-Flight Swap Balances in LiquidLaneAdapter) was set aside because its code citation does not support it. 81 were not checked for lack of check budget, 16 because no matching code was found, 4 because the audited repository was unavailable, 2 because the file path differs from the audit scope and 1 because the audit names no contract; 7 are gas-only findings with no security weight.
  • 20 of the 35 audit reports are hosted on GitHub rather than on the auditor's own site. Their provenance is unverified, which lowers confidence for claims that rest only on them. It is not a score penalty.
  • Deployed code was compared with audited code only in part. No bounded diff files were supplied for the rows that differ, so the materiality judgment below rests on file counts.
  • No collection gaps are listed. There are 94 inventory gaps: audit-named contracts with no attributed deployed address. The deployment inventory is therefore partial.
  • CertiK Skynet lists a score as a dated third-party snapshot . It is not a code audit and sets no level.

Audit Coverage

The protocol has many audits by independent firms across its product generations. Coverage of the deployed contracts is much thinner than the audit count suggests, because audited revisions are rarely tied to deployed addresses.

Audits in the workspace (35 reports; three of them are ChainSecurity auditor-page summaries of Core, Burners and Default Rewards that add no separate review). Publication provenance is unverified for 20 of them, see Data Completeness. Statemind is listed below but is not counted as an independent firm: the team biography says Symbiotic's co-founder is also Statemind's co-founder and CEO .

AuditorScopeDateFindings
ChainSecurityDefault Collateral token2024-05-142
StatemindDefault Collateral (legacy vaults)2024-04-26 to 2024-05-014
ZellicCore2024-07-303
ChainSecurityCore (also a listing page)2024-08-1313
ChainSecurityCollateral Migrator2024-08-161
CertoraCore, with formal verification2024-0815
OtterSecCore and Rewards2024-093
ChainSecurityBurners2024-11-1117
StatemindCore2024-06-03 to 2024-12-2035
CantinaCore competitionreport dated 2024-12-094
StatemindHooks2024-12-22 to 2024-12-277
ChainSecurityDefault Rewards2025-04-0710
StatemindMiddleware codebase at commit 5c801a0, AI-generated report2025-05-30117
BailsecVanguard pre-audit2025-06-1111
SherlockRelay contracts and network2025-06-19 to 2025-07-108
CyfrinRelay2025-09-1513
Sigma PrimeRelay contracts2025-0916
BailsecRewards2025-1130
CyfrinKeyRegistry2025-12-031
CyfrinBLS12-3812025-12-113
AckeeRewards V2 (final report 2025-12-21, fix review January 2026)2025-12-2110
StatemindCore V22026-05-28 to 2026-06-2641
BailsecVaults V22026-0660
BailsecRestaking Adapter (Core V2 extension)2026-0615
BailsecLiquid Lane adapters and Reactor2026-0646
OpenZeppelinCore v2 adapters (Liquid Lane)2026-06-2928
StatemindCore V2 ThreeFAdapter2026-06-30 to 2026-07-097
BailsecAsync/Account scopes 1 to 4 (mGLOBAL, 3F, Centrifuge, Figure and OpenEden)2026-0710, 12, 14, 16
BailsecEuler / Pareto2026-0814

The docs list the audits and the bug bounty on one page . The docs list a Cantina bug bounty at $500,000 for Core and Rewards . The Cantina program page could not be read, so its current status, its maximum payout for a critical finding and its scope list are not established. A LinkedIn post states a maximum reward of $500,000 (source reliability: unverified publisher) ; that post is undated and does not state the critical-tier amount.

Certora combined the Prover with manual review . The Cantina competition on core ran from Sep 11th to Oct 2nd and found 39 issues .

Unaudited or weakly covered areas:

  • The Statemind Middleware report is AI-generated and covers the Middleware codebase at commit 5c801a0 (the code check matched it to symbioticfi/relay-contracts); . Later human reviews by Sherlock, Cyfrin and Sigma Prime cover the relay contracts.
  • Per-vault curator and adapter factories (for example the AccountFactory contracts of Liquid Lane) are named in audit scopes, but the audited revision was not identified.
  • The audit publication hosts for many reports are not auditor-controlled, see Data Completeness.

Deployed code vs audits

Only 1 of 25 deployed component rows is shown identical to an audited version; 5 differ from the audited version, 17 are named in an audit scope but have no identified audited revision, and 2 have an audit scope that could not be determined.

Share of lines compared (protocol-owned verified deployed files, each file counted once): 7.3% of lines are identical to an audited version (1 file, 240 lines), 14.6% differ (9 files, 479 lines), 0.0% are cosmetic differences, 0.3% are unaudited (1 file, 9 lines) and 77.8% are unknown (33 files, 2546 lines) because the audited version was not identified. Changed, unknown and unaudited lines by reviewed criticality: critical 1135, high 956, peripheral 31, unrated 912 (roles: funds, access_control, upgradeability, oracle, view_util). Some code has not been assessed for criticality (912 unrated lines). No files lack line counts and no components are unmeasured. A difference in lines is not a finding of vulnerability, and an identical file is not proof of safety.

For the three critical rows that differ (the two MigratablesFactory rows and VaultFactory), the audited commit is symbioticfi/core@788f2d7 . These changes are judged material because no diff supports a cosmetic reading.

ComponentDeploymentStatusAudit and revisionCriticalityNote
AA_FalconX_AccountFactoryeip155:1:0x161954842b7ea47cbd050cab4875daa4d6599476<br>eip155:1:0x2fb3b8321951954b18ce5bf06da1cf5a6363bc0a<br>eip155:1:0x58f61759d858eafd8f58e4926e251701f6495dbf<br>eip155:1:0x6f80fd397120164e44d23b9334910c8de41ce4ce<br>eip155:1:0xb5eff01403c0dd3ada1010919791d6eb82f6a84c (6 total, first 5 shown; Sourcify runtime match)Audit scope could not be determined from the available materialnone identifiedcriticalVersion not compared
mGLO_Accounteip155:1:0x683f353ff35bee45d0a72acc24e65eff894d35cf (1 total; Sourcify runtime match)Audit scope could not be determined from the available materialnone identifiedcriticalVersion not compared
MigratablesFactoryeip155:1:0x3275ae068f4951e2e4d3dca107a54e4c219b02e7<br>eip155:1:0x70bc72b19a554436459a2c6a9e88892aed18685b<br>eip155:1:0x98a6aac642e8616514e19be37557c08e77fccbd8 (3 total; Etherscan lists verified source (explorer claim, not a runtime match))Differs from the audited versionOtterSec Core and Rewards audit (September 2024), symbioticfi/core@788f2d7criticalDiffers from the audited version in 3 files (plus 1 file not in the audited version); judged material (no supplied diff supports a cosmetic reading)
MigratablesFactoryeip155:1:0xc3398bfe441ef91bad0650116bc9a11ecaf843a5 (1 total; Etherscan lists verified source (explorer claim, not a runtime match))Differs from the audited versionOtterSec Core and Rewards audit (September 2024), symbioticfi/core@788f2d7criticalDiffers from the audited version in 3 files; judged material (no supplied diff supports a cosmetic reading)
VaultFactoryeip155:1:0xaeb6bdd95c502390db8f52c8909f703e9af6a346 (1 total; Sourcify runtime match)Differs from the audited versionOtterSec Core and Rewards audit (September 2024), symbioticfi/core@788f2d7criticalDiffers from the audited version in 4 files; judged material (no supplied diff supports a cosmetic reading)
NetworkMiddlewareServiceeip155:1:0xd7dc9b366c027743d90761f71858bca83c6899ad (1 total; Sourcify runtime match)Differs from the audited versionOtterSec Core and Rewards audit (September 2024), symbioticfi/core@788f2d7high (access_control)Differs from the audited version in 1 file
NetworkRegistryeip155:1:0xc773b1011461e7314cf05f97d95aa8e92c1fd8aa (1 total; Sourcify runtime match)Differs from the audited versionOtterSec Core and Rewards audit (September 2024), symbioticfi/core@788f2d7high (access_control, view_util)Differs from the audited version in 2 files
AA_FalconX_AccountFactoryeip155:1:0x7cd61e3364a84b7d50f0a42c6eb937a1f05a0254<br>eip155:1:0xa47ce86c304e251198f1a318c569e5217b694e1f (2 total; Etherscan lists verified source (explorer claim, not a runtime match))Named in an audit scope; audited revision not identifiedBailsec Euler/Pareto audit (2026-08), revision not identifiedcriticalVersion not compared: no confirmed audited revision
mROX_AccountFactoryeip155:1:0x60d9961702392fedf0c511ef2ef52834f50f1c1d (1 total; Etherscan lists verified source (explorer claim, not a runtime match))Named in an audit scope; audited revision not identifiedBailsec Euler/Pareto audit (2026-08), revision not identifiedcriticalVersion not compared: no confirmed audited revision
mM1USD_AccountFactoryeip155:1:0x45e234291f2d598813f160dfe0a089651b91db4b (1 total; Etherscan lists verified source (explorer claim, not a runtime match))Named in an audit scope; audited revision not identifiedBailsec Euler/Pareto audit (2026-08), revision not identifiedcriticalVersion not compared: no confirmed audited revision
ProtocolFeeRegistryeip155:1:0x9983edcd02f57b8b94bd8de9d18e614dfb964d6d (1 total; Sourcify runtime match)Named in an audit scope; audited revision not identifiedStatemind Core V2 audit (2026-05-28 to 2026-06-26), revision not identifiedcritical (funds, access_control, view_util)Version not compared: no confirmed audited revision
deJAAA_AccountFactoryeip155:1:0x85661e24b6843adcbf956891d52e519679a4ffb9 (1 total; Etherscan lists verified source (explorer claim, not a runtime match))Named in an audit scope; audited revision not identifiedBailsec Euler/Pareto audit (2026-08), revision not identifiedcriticalVersion not compared: no confirmed audited revision
deJTRSY_AccountFactoryeip155:1:0x0b8c1757d058edd122c3b1178ffa29637d825c61 (1 total; Etherscan lists verified source (explorer claim, not a runtime match))Named in an audit scope; audited revision not identifiedBailsec Euler/Pareto audit (2026-08), revision not identifiedcriticalVersion not compared: no confirmed audited revision
JAAA_AccountFactoryeip155:1:0xff7a0730b69c6d5bc0fb67879f31ae69b9cb76b6 (1 total; Etherscan lists verified source (explorer claim, not a runtime match))Named in an audit scope; audited revision not identifiedBailsec Euler/Pareto audit (2026-08), revision not identifiedcriticalVersion not compared: no confirmed audited revision
CarryTradeUSDTRYLeverage_AccountFactoryeip155:1:0x4cf6438a32c87e42a31b369f185d1d9f4e3b9d5b (1 total; Etherscan lists verified source (explorer claim, not a runtime match))Named in an audit scope; audited revision not identifiedBailsec Euler/Pareto audit (2026-08), revision not identifiedcriticalVersion not compared: no confirmed audited revision
mGLOBAL_AccountFactoryeip155:1:0xe2d6c6c2b715ce44bbeb001d16cf98610d7b102d (1 total; Etherscan lists verified source (explorer claim, not a runtime match))Named in an audit scope; audited revision not identifiedBailsec Euler/Pareto audit (2026-08), revision not identifiedcriticalVersion not compared: no confirmed audited revision
mHYPER_AccountFactoryeip155:1:0x88715f8366ea9481bca1882d79e0ef8b18f3a379 (1 total; Etherscan lists verified source (explorer claim, not a runtime match))Named in an audit scope; audited revision not identifiedBailsec Euler/Pareto audit (2026-08), revision not identifiedcriticalVersion not compared: no confirmed audited revision
HYB_AccountFactoryeip155:1:0x1c9f7e96631f7a44d63607eec7b5083f12b3c6cd (1 total; Etherscan lists verified source (explorer claim, not a runtime match))Named in an audit scope; audited revision not identifiedBailsec Euler/Pareto audit (2026-08), revision not identifiedcriticalVersion not compared: no confirmed audited revision
StockMarketTRBasisTrade_AccountFactoryeip155:1:0x4bf0a789a9b51a5947378c294b25012f5382a818 (1 total; Etherscan lists verified source (explorer claim, not a runtime match))Named in an audit scope; audited revision not identifiedBailsec Euler/Pareto audit (2026-08), revision not identifiedcriticalVersion not compared: no confirmed audited revision
JTRSY_AccountFactoryeip155:1:0xbeb8c94d05aea90263de6b54724e0fd4ef0fd96d (1 total; Etherscan lists verified source (explorer claim, not a runtime match))Named in an audit scope; audited revision not identifiedBailsec Euler/Pareto audit (2026-08), revision not identifiedcriticalVersion not compared: no confirmed audited revision
mFONE_AccountFactoryeip155:1:0x0eb4c55ba5bf3cd73eaa76330122ba59349e3fe6 (1 total; Etherscan lists verified source (explorer claim, not a runtime match))Named in an audit scope; audited revision not identifiedBailsec Euler/Pareto audit (2026-08), revision not identifiedcriticalVersion not compared: no confirmed audited revision
PRIME_AccountFactoryeip155:1:0xe73d29fb92340a16625388712a93490e159e433e (1 total; Etherscan lists verified source (explorer claim, not a runtime match))Named in an audit scope; audited revision not identifiedBailsec Euler/Pareto audit (2026-08), revision not identifiedcriticalVersion not compared: no confirmed audited revision
AccountRegistryeip155:1:0x29f57ed0f855f18aec82146a0b7f15cab66a1c8d (1 total; Etherscan lists verified source (explorer claim, not a runtime match))Named in an audit scope; audited revision not identifiedOpenZeppelin Core v2 adapters audit (2026-06-29), revision not identifiedhigh (access_control)Version not compared: no confirmed audited revision
AdapterRegistryeip155:1:0x788823c9579a049d986eb76718bc758c0998594a (1 total; Sourcify runtime match)Named in an audit scope; audited revision not identifiedStatemind Core V2 audit (2026-05-28 to 2026-06-26), revision not identifiedhighVersion not compared: no confirmed audited revision
Reactoreip155:1:0xc323b898d7e4105e3980082b74cc5d4602996b10 (1 total; Sourcify runtime match)Identical to the audited versionBailsec Liquid Lane adapters and Reactor audit (2026-06), symbioticfi/rfq@041aab4None ()The protocol-owned source files Sourcify matches at this address are identical to symbioticfi/rfq@041aab4 audited by Bailsec (library files not compared)
177 further names from audit scopes and findings (for example VaultV2, Settlement, Rewards, BurnerRouter)no attributed deployed addressNamed by audits; no current deployed address attributed to themvarious auditscritical, high or peripheral per componentNot compared; deployed code for these names was not identified

The deployment inventory is partial: 94 audit-named contracts have no attributed current address (for example the relay-contracts Settlement, KeyRegistry and SigVerifierBlsBn254ZK, and VaultV2). The map lists no collection gaps. Criticality was rated for all 197 components and names that required it; 117 are critical, 32 high and 48 peripheral.

Findings by Severity

All lines below come from the audit findings not marked fixed. The deciding code check for every line is from the latest public repository (no deployed-code comparison exists for these findings). Where several audits report the same issue, the line names each audit. The four critical findings and most high findings come from the AI-generated Statemind Middleware report (Statemind is affiliated with Symbiotic's co-founder); the Sherlock, Cyfrin and Sigma Prime relay reviews are later human reviews of the same product area. Eleven high, twenty-four medium and one critical group are marked fixed and are not re-checked here.

Critical

  • Missing subgroup checks for BLS cryptographic inputs (Statemind Middleware report (AI-generated, 2025-05-30); member findings: "Missing Subgroup Checks for BLS Cryptographic Inputs"; "Missing Subgroup Check for BLS Keys") — — Audit status: open — Check: the latest public repository code (symbioticfi/relay-contracts@668d587) still contains the code this finding describes; deployed code was not compared (latest repository)
  • Incorrect generalized index used for SSZ proof verification (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: the audited file was not found at its audited path
  • Missing previous header hash consistency check in setValSetHeader (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: no matching code was located in the repository code
  • Unauthenticated key setters allow key registration for arbitrary operators (Statemind Middleware report (AI-generated, 2025-05-30); member findings: "Public Functions Allow Unauthenticated Key Registration for Arbitrary Operators"; "Direct Calls to setKey32/setKey64 Bypass Critical Validation and State Updates"; "Direct Call to setKey(operator, keyTag, key) Overload Bypasses Signature Verification and Key Management Logic"; "State Inconsistency from Unauthenticated Key Setters") — — Audit status: open — Check: not checked: no matching code was located in the repository code

High

  • Incorrect proof height used in verifyKeyTagLocal (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: the audited file was not found at its audited path
  • Incorrect negation of G1 points with Y-coordinate equal to zero (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: the latest public repository code (symbioticfi/relay-contracts@668d587) still contains the code this finding describes; deployed code was not compared (latest repository)
  • Dependency on unaudited SqrtMod implementation (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: no matching code was located in the repository code
  • Non-canonical point coordinates used in challenge hash calculation (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: the latest public repository code (symbioticfi/relay-contracts@668d587) still contains the code this finding describes; deployed code was not compared (latest repository)
  • Initializer can be front-run leading to malicious configuration (Statemind Middleware report (AI-generated, 2025-05-30); member findings: "Initializer Can Be Front-Run Leading to Malicious Configuration"; "Initialization Can Be Front-Run with Attacker-Controlled Parameters") — — Audit status: open — Check: not checked: no matching code was located in the repository code; not checked: check budget
  • Incorrect invariant check in setEpochDuration can lead to epochDuration <= commitDuration (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: no matching code was located in the repository code
  • Timestamp calculation overflow leads to incorrect epoch start times (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: no matching code was located in the repository code
  • Constructor validates state variables that are empty at construction (Statemind Middleware report (AI-generated, 2025-05-30); member findings: "Constructor Reads Uninitialized State Variables for Input Validation"; "Unvalidated Input Array Lengths in Constructor Leads to Potential Out-of-Bounds Read") — — Audit status: open — Check: a fix is visible in the latest repository code (symbioticfi/relay-contracts@668d587); deployment not confirmed (latest repository)
  • Potential division by zero if epoch duration is set to zero (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: a fix is visible in the latest repository code (symbioticfi/relay-contracts@668d587); deployment not confirmed (latest repository)
  • Base contract initializers are not called before Settlement-specific initialization (Statemind Middleware report (AI-generated, 2025-05-30); member findings: "Incomplete Initialization of Inherited Base Contracts in __Settlement_init"; "Base contract initializers must be called before Settlement-specific initialization") — — Audit status: open — Check: the code was rewritten; the check could not determine whether this finding still applies (latest repository)
  • keyTag not validated against required key tag for the epoch (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: the latest public repository code (symbioticfi/relay-contracts@668d587) may still contain the code this finding describes; the check verdict is issue present with confidence 5, below the level for a firm statement, deployed code was not compared (latest repository)
  • Loop bound mismatch with decoded array lengths in SigVerifierBlsBn254Simple (Statemind Middleware report (AI-generated, 2025-05-30); member findings: "Loop Bound Mismatch with Decoded Array Lengths in SigVerifierBlsBn254Simple"; "Decoded ValidatorData array length mismatch with proof's specified length field") — — Audit status: open — Check: the code was rewritten; the check could not determine whether this finding still applies (latest repository)
  • Unclear message hashing responsibility and method for EdDSA verification (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: no matching code was located in the repository code
  • Incorrect logic in getCurrentValue leads to wrong epoch data retrieval (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: no matching code was located in the repository code
  • Missing events for critical state and parameter changes (Statemind Middleware report (AI-generated, 2025-05-30); member findings: "Missing Events for Critical State Changes"; "Missing events for critical parameter changes and header commits"; "Critical Parameter Change Lacks Event Emission") — — Audit status: open — Check: not checked: no matching code was located in the repository code; not checked: check budget; not checked: no code-check row for this member
  • Failure in unregistration logic prevents force-pausing (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: no matching code was located in the repository code
  • Misleading KeyTags duplicate revert for invalid key tags in serialize (Statemind Middleware report (AI-generated, 2025-05-30); member findings: "Misleading KeyTags_Duplicate Revert for Invalid Key Tags in serialize Function"; "Misleading KeyTags Duplicate Revert for Invalid Key Tags in serialize Function") — — Audit status: open — Check: not checked: the finding names no contract; not checked: no code-check row for this member
  • Potential gas griefing/DoS via unbounded iteration in value retrieval functions (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: gas/efficiency finding; no security weight
  • Message parameter assumed to be a 32-byte hash without an explicit type (Statemind Middleware report (AI-generated, 2025-05-30); member findings: "Ambiguous message Parameter Assumed to be a 32-Byte Hash"; "Implicit 32-Byte Length Requirement for message Parameter"; "Implicit Assumption on message Length for abi.decode") — — Audit status: open — Check: the latest public repository code (symbioticfi/relay-contracts@668d587) still contains the code this finding describes; deployed code was not compared (latest repository)
  • Element duplication and state corruption if initial addition key is zero (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: a fix is visible in the latest repository code (symbioticfi/relay-contracts@668d587); deployment not confirmed (latest repository)
  • EdDSA public keys lack identity and subgroup checks (Statemind Middleware report (AI-generated, 2025-05-30); member findings: "EdDSA Verification Does Not Explicitly Reject Identity Public Key"; "Missing Subgroup and Identity Checks for Ed25519 Public Keys") — — Audit status: open — Check: not checked: no matching code was located in the repository code
  • Governance Privilege (Bailsec Rewards audit (2025-11)) — severity High (estimated): AI estimate: the audit states no numeric rating, only a governance label for privileges that can lead to loss of funds, which maps to high; the audit text is cited for the issue, not for the rating — — Audit status: acknowledged — Check: not checked: the audited repository code was not available
  • Cross-contract replay attack due to hardcoded version within OzEIP712 (Bailsec Rewards audit (2025-11)) — — Audit status: acknowledged — Check: not checked: the audited repository code was not available
  • Zero initial depositor can permanently brick whitelisted deposits (Bailsec Vaults V2 audit (2026-06)) — — Audit status: acknowledged — Check: not checked: the audited repository code was not available
  • Vault composition can be overridden via donation attacks (Bailsec Vaults V2 audit (2026-06); member findings: "Vault composition can be overridden via donation attacks"; "Vault is susceptible to inflation attacks via donations") — — Audit status: partially fixed — Check: not checked: the audited repository code was not available; not checked: check budget

Medium

  • Incorrect amount of slashing and too early execution of slashing (Statemind Core audit (2024-06-03 to 2024-12-20)) — — Audit status: acknowledged — Check: not checked: check budget
  • Vault can be DoS-ed for deposits after a few slashing rounds (Certora Core audit with formal verification (2024-08)) — — Audit status: acknowledged — Check: not checked: check budget
  • Factory approved implementation cannot be blocked (Certora Core audit with formal verification (2024-08)) — — Audit status: partially fixed — Check: not checked: check budget
  • claimBatch can revert mid batch, blocking the entire batch from completing (Certora Core audit with formal verification (2024-08)) — — Audit status: acknowledged — Check: not checked: check budget
  • Slashable amounts are calculated wrong and it is possible for depositors to abuse it (Cantina Core competition report (2024-12-09)) — — Audit status: not stated — Check: not checked: check budget
  • Vetoslashing might fail (Cantina Core competition report (2024-12-09)) — — Audit status: partially fixed — Check: not checked: check budget
  • Potential Over-Slashing in Multi-Operator and Multi-Subnetwork Scenarios (Cantina Core competition report (2024-12-09)) — — Audit status: acknowledged — Check: not checked: check budget
  • _remove with key = 0 can revert due to checkpoint key order violation (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: check budget
  • deserialize accepts non-canonical keys and coordinates equal to the field modulus (Statemind Middleware report (AI-generated, 2025-05-30); member findings: "Incomplete Validation in deserialize Allows Non-Canonical Keys"; "deserialize Can Produce Keys with Y-Coordinate Equal to Field Modulus") — — Audit status: open — Check: not checked: check budget
  • Identity point accepted for G2 key component (keyG2) (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: check budget
  • Gas griefing vulnerability due to unbounded isNonSigners array decoding (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: gas/efficiency finding; no security weight
  • Missing input validation for proof calldata length (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: check budget
  • Getter functions returning unbounded arrays may cause denial of service (Statemind Middleware report (AI-generated, 2025-05-30); member findings: "Potentially Unbounded Loops in Getter Functions May Lead to Denial of Service"; "Getter functions returning unbounded arrays may lead to denial of service"; "Public Getter Functions Returning Potentially Unbounded Arrays May Lead to Denial of Service"; "Potential Denial of Service from Unbounded Array Returns in Basic List Getter Functions") — — Audit status: open — Check: not checked: check budget
  • setCommitDuration does not validate against epoch duration invariant (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: check budget
  • registerOperator susceptible to malicious OPERATOR_REGISTRY if misused (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: check budget
  • Inefficient memory allocation and iteration in _valuesAt and _values (Statemind Middleware report (AI-generated, 2025-05-30); member findings: "Inefficient Memory Allocation and Iteration in _valuesAt and _values"; "Gas Intensive Default Hint Allocation in _valuesAt"; "Inefficient memory allocation and iteration in _valuesAt/_values for high-churn sets") — — Audit status: open — Check: not checked: no code-check row for this member
  • Ed25519 decoding does not reject y >= p (Statemind Middleware report (AI-generated, 2025-05-30); member findings: "Non-compliance with RFC 8032 Ed25519 Decoding: y >= p Check Missing"; "Non-compliance with RFC 8032 Ed25519 decoding: missing y >= p check") — — Audit status: open — Check: not checked: check budget; not checked: no code-check row for this member
  • Unvalidated G2 public key component from proof passed to verification (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: check budget
  • No relational validation between minInclusionVotingPower and maxVotingPower (Statemind Middleware report (AI-generated, 2025-05-30); member findings: "Lack of Relational Validation Between minInclusionVotingPower and maxVotingPower"; "No relational validation between minInclusionVotingPower and maxVotingPower") — — Audit status: open — Check: not checked: check budget
  • Prolong window uses current commitDuration instead of the historical value (Statemind Middleware report (AI-generated, 2025-05-30); member findings: "Prolong Window Calculation Uses Current commitDuration Instead of Relevant Historical Value"; "Prolong window calculation uses current commitDuration instead of historical value") — — Audit status: open — Check: not checked: check budget
  • Functions without explicit role assignment default to DEFAULT_ADMIN_ROLE (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: check budget
  • Missing events for shared vault registration and unregistration (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: check budget
  • Identity point acceptance for signature (signatureG1) (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: check budget
  • Key collision for indexed getKey with index 0 (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: check budget
  • processInclusionProofSha256 incorrectly handles proofs of height zero (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: check budget
  • Potential signature bypass via external hashTypedDataV4 function pointer (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: check budget
  • Unregistration failure prevents unwhitelisting (Statemind Middleware report (AI-generated, 2025-05-30); member findings: "Unwhitelisting can fail if underlying unregistration reverts"; "Failure in unregistration logic prevents unwhitelisting") — — Audit status: open — Check: not checked: check budget
  • Potential division by zero in decompression if denominator is zero (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: check budget
  • Slashing window can only be decreased and may become permanently set (Statemind Middleware report (AI-generated, 2025-05-30); member findings: "Slashing window parameter can only be decreased and may become permanently zero"; "Slashing window can only be decreased and potentially permanently set to zero") — — Audit status: open — Check: not checked: check budget
  • Denial of service via memory allocation in getKeysAt due to empty hints (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: check budget
  • Gas inefficiency in EIP-712 hash computations (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: gas/efficiency finding; no security weight
  • deserialize does not verify reconstructed point satisfies curve equation (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: check budget
  • Ed25519 decoding does not reject x=0 with sign bit set (Statemind Middleware report (AI-generated, 2025-05-30); member findings: "Non-compliance with RFC 8032 Ed25519 Decoding: Failure to Reject (x=0, sign_bit=1) Keys"; "Non-compliance with RFC 8032 Ed25519 Decoding: Failure to Reject (x=0, sign bit=1) Keys") — — Audit status: open — Check: not checked: check budget; not checked: no code-check row for this member
  • Potential revert in _add when re-adding with out-of-order key (Statemind Middleware report (AI-generated, 2025-05-30)) — — Audit status: open — Check: not checked: check budget
  • epochDuration check is incorrect when comparing with slashingWindow (Bailsec Vanguard pre-audit (2025-06-11)) — — Audit status: acknowledged — Check: not checked: check budget
  • Attacker will manipulate voting power calculations as getOperatorVotingPower() (Sherlock relay contracts review (2025-06-19 to 2025-07-10)) — — Audit status: acknowledged — Check: the latest public repository code (symbioticfi/relay-contracts@668d587) still contains the code this finding describes; deployed code was not compared (latest repository)
  • Most KeyRegistry, VotingPowerProvider functions can be DoSed (Sherlock relay contracts review (2025-06-19 to 2025-07-10)) — — Audit status: acknowledged — Check: the latest public repository code (symbioticfi/relay-contracts@668d587) still contains the code this finding describes; deployed code was not compared (latest repository)
  • Changing the epoch duration will completely break the vault and the slashers (Sherlock relay contracts review (2025-06-19 to 2025-07-10)) — — Audit status: acknowledged — Check: the latest public repository code (symbioticfi/relay-contracts@668d587) may still contain the code this finding describes; low confidence in the match, deployed code was not compared (latest repository)
  • A malicious operator will control consensus without risking stake (stake-exit attack) (Sherlock relay contracts review (2025-06-19 to 2025-07-10)) — — Audit status: acknowledged — Check: the latest public repository code (symbioticfi/relay-contracts@668d587) may still contain the code this finding describes; low confidence in the match, deployed code was not compared (latest repository)
  • Voting power returns 0 during oracle failures causing unfair voting outcomes (Cyfrin Relay audit (2025-09-15)) — — Audit status: acknowledged — Check: not checked: check budget
  • Precision loss causes systematic under-reporting of voting power (Cyfrin Relay audit (2025-09-15)) — — Audit status: acknowledged — Check: not checked: check budget
  • Missing Multi-Chain Voting Power Distribution Controls (Sigma Prime Relay Contracts audit (2025-09)) — — Audit status: acknowledged — Check: not checked: check budget
  • Symbiotic vaults are not guaranteed to have corresponding owners (Bailsec Rewards audit (2025-11)) — — Audit status: acknowledged — Check: not checked: check budget
  • Curator and operator fee will return zero if vault timestamp is before contract deployment timestamp (Bailsec Rewards audit (2025-11)) — — Audit status: acknowledged — Check: not checked: check budget
  • protocolFee ignores disabled default fee (reported in several audits as: "protocolFee ignores isEnabled flag in fallback scenario" (Bailsec Rewards audit (2025-11)); "protocolFee ignores disabled default fee" (Ackee Rewards V2 audit (2026-01))) — — Audit status: acknowledged — Check: not checked: check budget; not checked: no code-check row for this member
  • Race condition due to protocolFee increase (Bailsec Rewards audit (2025-11); member findings: "Race condition for protocolFee increase is amplified due to permissionless nature of distributeCumulativeMerkleRewards"; "Race condition due to protocolFee increase") — — Audit status: acknowledged — Check: not checked: check budget
  • Skippable rewards in VaultSnapshotRewards become unclaimable (reported in several audits as: "Footgun within claimVaultSnapshotRewards can result in lost rewards" (Bailsec Rewards audit (2025-11)); "Skippable rewards in VaultSnapshotRewards become unclaimable" (Ackee Rewards V2 audit (2026-01))) — — Audit status: acknowledged — Check: not checked: check budget; not checked: no matching code was located in the repository code
  • CowSwap convert can be sandwiched for profit (Bailsec Restaking Adapter audit (2026-06)) — — Audit status: acknowledged — Check: not checked: check budget
  • Restaking slashes are denominated in shares and can thus burn unpredictable amounts of assets (Bailsec Restaking Adapter audit (2026-06)) — — Audit status: acknowledged — Check: not checked: check budget
  • Merkle claims and converts can be sandwiched for profit (Bailsec Restaking Adapter audit (2026-06)) — — Audit status: acknowledged — Check: not checked: check budget
  • Permissionless delayed CoW conversion allows attacker-chosen limit prices (reported in several audits as: "Open convert function allows any swap price" (Bailsec Restaking Adapter audit (2026-06)); "Permissionless delayed CoW conversion allows attacker-chosen limit prices" (Bailsec Liquid Lane adapters and Reactor audit (2026-06))) — — Audit status: acknowledged — Check: not checked: no code-check row for this member; not checked: check budget
  • RFQ swap outputs can be routed away while Reactor pays from shared executor balances (Bailsec Liquid Lane adapters and Reactor audit (2026-06)) — — Audit status: acknowledged — Check: not checked: check budget
  • Linear Scan of the converters Array Increases Gas Cost of Conversions (reported in several audits as: "Unbounded converter list can make converter checks expensive" (Bailsec Liquid Lane adapters and Reactor audit (2026-06)); "Linear Scan of the converters Array Increases Gas Cost of Conversions" (OpenZeppelin Core v2 adapters audit (2026-06-29))) — — Audit status: acknowledged — Check: not checked: no code-check row for this member
  • Existing accounts keep stale owner and converter authority after adapter ownership transfer (Bailsec Liquid Lane adapters and Reactor audit (2026-06)) — — Audit status: acknowledged — Check: not checked: check budget
  • Stale filler approvals reactivate when a market maker address is reused (Bailsec Liquid Lane adapters and Reactor audit (2026-06)) — — Audit status: acknowledged — Check: not checked: check budget
  • Asymmetric State Handling Between Adding and Removing a Token-to-Redeem (reported in several audits as: "Token readd reuses stale per-token state" (Bailsec Liquid Lane adapters and Reactor audit (2026-06)); "Asymmetric State Handling Between Adding and Removing a Token-to-Redeem" (OpenZeppelin Core v2 adapters audit (2026-06-29))) — — Audit status: acknowledged — Check: not checked: check budget; not checked: no code-check row for this member
  • Direct token transfers affect accounting (reported in several audits as: "Donation attack via raw balanceOf accounting in LiquidLaneAdapter" (Bailsec Liquid Lane adapters and Reactor audit (2026-06)); "Direct token transfers affect accounting" (Bailsec Async/Account scope 1 audit (2026-07))) — — Audit status: acknowledged — Check: not checked: check budget; not checked: no code-check row for this member
  • Discount swap has no slippage protection on execution-time oracle pricing (Bailsec Liquid Lane adapters and Reactor audit (2026-06); member findings: "Discount swap has no slippage protection on execution-time oracle pricing"; "Discount swaps can be sandwiched for profit") — — Audit status: acknowledged — Check: not checked: check budget
  • Midas redemption fee not incorporated (Bailsec Liquid Lane adapters and Reactor audit (2026-06)) — — Audit status: acknowledged — Check: not checked: check budget
  • Account factory selection is not post-validated and can break LiquidLane account invariants (Bailsec Liquid Lane adapters and Reactor audit (2026-06)) — — Audit status: acknowledged — Check: not checked: check budget
  • Midas redemption allowances can strand submitted requests (Bailsec Liquid Lane adapters and Reactor audit (2026-06)) — — Audit status: acknowledged — Check: not checked: check budget
  • Pending Midas request marks can diverge from settlement rates (Bailsec Liquid Lane adapters and Reactor audit (2026-06)) — — Audit status: partially fixed — Check: not checked: check budget
  • Redemption token price not incorporated (Bailsec Liquid Lane adapters and Reactor audit (2026-06); member findings: "Fallback redemption tokens are counted at decimal par"; "Redemption token price not incorporated") — — Audit status: acknowledged — Check: not checked: check budget
  • Token-to-Redeem Balances Below the Redemption Vault Minimum Freeze sync and convert (reported in several audits as: "Midas dust can block the conversion of settled fallback redemption tokens" (Bailsec Liquid Lane adapters and Reactor audit (2026-06)); "Permissionless dust sync can delay redemptions" (Bailsec Liquid Lane adapters and Reactor audit (2026-06)); "Token-to-Redeem Balances Below the Redemption Vault Minimum Freeze sync and convert" (OpenZeppelin Core v2 adapters audit (2026-06-29))) — — Audit status: partially fixed — Check: the code was rewritten; the check could not determine whether this finding still applies (latest repository)
  • Deallocation Reverts on a Frozen Account, Bricking Vault-Wide Operations (reported in several audits as: "Deallocation Reverts on a Frozen Account, Bricking Vault-Wide Operations" (OpenZeppelin Core v2 adapters audit (2026-06-29)); "A single frozen or blacklisted redemption account freezes deallocation and all vault withdrawals" (Bailsec Liquid Lane adapters and Reactor audit (2026-06))) — — Audit status: acknowledged — Check: the latest public repository code (symbioticfi/core@25a065a) still contains the code this finding describes; deployed code was not compared (latest repository)
  • BTC and ETH mTokens Cannot Be Redeemed, Bricking Swaps and Stranding Inventory (reported in several audits as: "BTC and ETH mTokens Cannot Be Redeemed, Bricking Swaps and Stranding Inventory" (OpenZeppelin Core v2 adapters audit (2026-06-29)); "Immutable redemption token cannot track Midas payment token removal" (Bailsec Liquid Lane adapters and Reactor audit (2026-06))) — — Audit status: acknowledged — Check: the code was rewritten; the check could not determine whether this finding still applies (latest repository)
  • Single Token Oracle Failure Can Brick the Vault With Self-Locked Recovery (reported in several audits as: "Single Token Oracle Failure Can Brick the Vault With Self-Locked Recovery" (OpenZeppelin Core v2 adapters audit (2026-06-29)); "A reverting Midas data feed bricks all vault accounting and cannot be removed or repointed" (Bailsec Liquid Lane adapters and Reactor audit (2026-06))) — — Audit status: partially fixed — Check: the code was rewritten; the check could not determine whether this finding still applies (latest repository)
  • Pending queue can be griefed by any user (Bailsec Vaults V2 audit (2026-06)) — — Audit status: acknowledged — Check: not checked: check budget
  • Temporary deposits can bypass AppAdapter share caps (Bailsec Vaults V2 audit (2026-06)) — — Audit status: acknowledged — Check: not checked: check budget
  • Pending withdrawal debt can be assigned to adapters with no-op requestDeallocate (Bailsec Vaults V2 audit (2026-06)) — — Audit status: acknowledged — Check: not checked: check budget
  • sweepPending can be abused to selectively drain adapters (Bailsec Vaults V2 audit (2026-06)) — — Audit status: acknowledged — Check: not checked: check budget
  • Morpho liquidity adapter accounting assets can overstate executable withdrawal liquidity (Bailsec Vaults V2 audit (2026-06); member findings: "Morpho liquidity adapter accounting assets can overstate executable withdrawal liquidity"; "Morpho vault balance can overstate the deallocatable amount") — — Audit status: acknowledged — Check: not checked: check budget
  • In-Flight Swap Balances in LiquidLaneAdapter Can Be Stolen by Arbitrary Callers (OpenZeppelin Core v2 adapters audit (2026-06-29)) — — Audit status: acknowledged — Check: not checked: the repository match was set aside because the cited code does not support the finding; the swap entry points in the latest repository code (symbioticfi/core@25a065a) validate the caller against the owner, market maker or authorized fillers, so arbitrary access is not established (latest repository)
  • Predictable Jumps in totalAssets Allow Timed Value Extraction From Passive Holders (OpenZeppelin Core v2 adapters audit (2026-06-29)) — — Audit status: acknowledged — Check: the latest public repository code (symbioticfi/core@25a065a) may still contain the code this finding describes; low confidence in the match, deployed code was not compared (latest repository)
  • mGLOBAL sync can revert on sub-minimum balances or Midas access-control failures, blocking finalization and swaps (Bailsec Async/Account scope 1 audit (2026-07)) — — Audit status: acknowledged — Check: not checked: check budget
  • wYLDS tokens stranded in subaccounts on cancellation (Bailsec Async/Account scope 4 audit (2026-07)) — — Audit status: acknowledged — Check: not checked: check budget
  • sync can revert on sub-minimum balances blocking finalization and swaps (Bailsec Async/Account scope 4 audit (2026-07)) — — Audit status: acknowledged — Check: not checked: check budget
  • Pareto default can leave AA FalconX priced after withdrawals stop (Bailsec Euler/Pareto audit (2026-08)) — — Audit status: partially fixed — Check: not checked: check budget
  • Yield recognized as a discrete step at request maturity lets zero-duration (JIT) depositors snipe other depositors' realized 3F yield (reported in several audits as: "totalAssets jumps when canWithdraw flips" (Bailsec Async/Account scope 2 audit (2026-07)); "Yield recognized as a discrete step at request maturity lets zero-duration (JIT) depositors snipe other depositors' realized 3F yield" (Statemind Core V2 ThreeFAdapter audit (2026-07-09))) — — Audit status: acknowledged — Check: not checked: check budget; not checked: no code-check row for this member

Low / Informational

The findings list holds 125 low-severity groups (101 not marked fixed, 24 fixed) and 251 informational groups (175 not marked fixed, 76 fixed). They carry no score weight here and are not itemised. Their dominant statuses are acknowledged (95 low, 131 informational) and open (33 informational).

Recurring Patterns

Eleven unresolved groups of medium or higher severity are reported by more than one audit; all are in the lists above. The clearest repeats are:

  • Liquid Lane account and redemption handling: token-to-redeem state, direct token transfers affecting accounting, deallocation reverting on a frozen account, oracle failure bricking the vault, and BTC/ETH mToken redemption. Bailsec (2026-06) and OpenZeppelin (2026-06-29), two independent firms, both report these.
  • Reward fee handling: the protocolFee ignoring a disabled default fee and skippable rewards becoming unclaimable, reported by Bailsec (2025-11) and Ackee (2025-12-21).
  • CoW conversion pricing and sandwiching, reported by Bailsec's Restaking Adapter and Liquid Lane audits.
  • Zero-duration deposit yield capture (just-in-time deposits), reported by Bailsec (Async/Account scope 2, 2026-07) and Statemind (ThreeFAdapter, 2026-07-09).
  • Smaller overlaps between Statemind Core V2 and Bailsec Vaults V2 (adapter removal blockable by dust, adapter deallocation return values, instant-capacity overstatement); these are low or informational and mostly acknowledged.

Code Quality Assessment

Assessment: audit effort is large and recent, but the code base moves quickly, and traceability from audit to deployment is weak. The basis is that the latest audits date from June to August 2026, 17 deployed component rows have no identified audited revision and five differ from the audited commit.

  • Review depth: independent firms reviewed the version-1 core (Zellic, ChainSecurity, Certora, OtterSec, a Cantina competition) and the version-2 vaults and adapters (Bailsec, OpenZeppelin; Ackee reviewed Rewards V2). Statemind also reviewed both versions, but it is affiliated with Symbiotic's co-founder, so its reviews add no independent weight. The Certora formal verification was performed on Core in August 2024.
  • Fix rate: of 520 issue groups, 76 informational, 24 low, 24 medium, 11 high and 1 critical are marked fixed. Acknowledged is the most common non-fixed status, so many findings are accepted design trade-offs, not code changes.
  • Open critical items: all four come from the AI-generated Statemind report. Only one was matched to code: the code check found the described missing BLS subgroup checks still in the latest relay-contracts repository (confidence 7). The other three critical items (SSZ proof index, previous header hash check, unauthenticated key setters) were not located or not checked. Among high-severity items, negation of points with Y equal to zero (confidence 7), non-canonical coordinates (confidence 6) and the message-as-32-byte-hash design item (confidence 6) also still appear. Sherlock's relay findings that still appear in the latest repository are listed under Medium.
  • Unchecked volume: 81 selected findings were not checked for lack of check budget. The unresolved medium findings in the Liquid Lane and Vaults V2 areas therefore rest on audit status alone.
  • Operational assurance: the docs list a Cantina bounty of $500,000 for "Core & Rewards" . Its current status, critical-tier payout and scope list are not established, and whether Liquid Lane, relay or Vaults V2 code is in scope is unknown.

Score: 4.3/10

Rubric security-score-v2.5; data completeness 1.00

Code security factors, points out of 10 (points)
Audit coverage
5.50 points
Audit depth
7.50 points
Unresolved findings
0.00 points
Assurance
5.50 points
Audit findings not marked fixed (findings)
Critical · open
4 findings
High · open
21 findings
High · partially fixed
1 findings
High · acknowledged
3 findings
Medium · open
27 findings
Medium · not stated
1 findings
Medium · partially fixed
6 findings
Medium · acknowledged
45 findings
FactorLevelEvidenceWeightPoints
Audit coverage (computed)L2supported0.355.50
Audit depthL3supported0.207.50
Unresolved findings (computed)L0supported0.300.00
AssuranceL2supported0.155.50

140 citations · click markers to view evidence

💰

Economic Security

5.3/10

Economic Security Analysis: Symbiotic

Symbiotic is an Ethereum-only collateral-markets protocol (about USD 317M TVL) where depositor funds sit in curator-run vaults that can be slashed by applications or lent out through adapters. The core vault accounting is audited and the sources record no bad debt or hack, but the economic design leaves depositors exposed to curator choices, third-party strategy layers (Morpho, Aave, Euler, Pareto, Midas and similar) and accounting edge cases that auditors left open. The sources identify no insurance or reserve fund. The one completed DEX probe, of the legacy DC_wstETH wrapper, shows no depth, though that wrapper redeems directly for its underlying token. Confidence is moderate: mechanism documents and audits are rich, but holder concentration, yield split, and bad-debt records are not published.

TVL & Liquidity Profile

  • Latest snapshot. DeFiLlama records (collected 2026-10-29 05:08 UTC) in the category "Collateral Markets". All of it is on Ethereum . No other chain holds recorded TVL. Bridge exposure of the deposited tokens themselves is a separate question (see Economic Attack Vectors).
  • Chain profile. The chain profile lists only Ethereum, with no L2BEAT stage or exit window, because it is the base chain .
  • Collateral mix (DefiLlama composition, 2026-10-29). Roughly 30% of TVL is uniBTC and 26% wstETH. Another 15% is LBTC. A BTC-wrapper concentration therefore sits behind nearly half of the TVL.
Symbiotic TVL by deposited token, USD m, DefiLlama composition 2026-10-29 (USD m)
uniBTC
96.29 USD m
wstETH
81.75 USD m
LBTC
46.61 USD m
rETH
21.76 USD m
weETH
19.29 USD m
SPK
13.26 USD m
  • Exit depth (measured). Users hold vault shares and the legacy DefaultCollateral wrapper tokens. The wrapper burns its receipt for an equal amount of the backing asset , so a thin wrapper pool does not by itself trap holders. The only wrapper with a completed DEX probe is DC_wstETH: a USD 100k probe moved the price by about 98% and no venue was found . The measured 2% depth is zero . The 0.5% depth (a coarse-probe lower bound) is null, which is a gap, not zero. The probes for DC_rETH and DC_sfrxETH returned errors, so those are unmeasured . No DEX depth is measured for the tokenized Core V2 vault shares.
  • Native exit. Depositors exit through the vault, not a DEX. Legacy (V1) vaults use an epoch-based request-and-claim process . Core V2 vaults add a withdrawal queue and adapter liquidity dependencies (see Solvency below). The Liquid Lane documentation says vaults "on average have a 14 day lockup period" .
  • Depositor concentration. Top-10 depositor or holder shares are not published in the sources, and no separate holder-distribution check was run. This is a gap. The vault listing shows large single vaults, for example a uniBTC vault with "Tvl 67.45M" .
  • Campaign dependence. Core V2 vaults earn points on assets allocated to adapters . The share of TVL tied to the points campaign is not cited anywhere, so it is not assessed.

Oracle Risk

  • Scope. The core vault, delegator and slasher contracts do not price assets. Slashing is triggered by an application's request and checked only for timing and size . Oracles enter through the Core V2 Liquid Lane adapters, which value redemption-asset holdings (Midas, Pareto, Centrifuge, OpenEden and Hastra tokens) for share pricing and instant swaps.
  • Midas feed. MidasOracle forwards a price from an external data feed with a hard min/max bound. On-chain at Ethereum block 26296259 (2026-10-29), the mGLO oracle shows . A Bailsec audit (2026) says the oracle now enforces the bound but "stale in-range prices still pass" . The same finding says the oracle accepts any price the external feed accepts . The bound is narrow protection; freshness and deviation checks are not enforced locally.
  • Pareto. A Bailsec audit found the oracle prices AA_FalconX only from the Pareto vault virtual price, without checking for default . The fix is partial: the defaulted token is valued at zero, but intake stays enabled .
  • Single-feed failure. An OpenZeppelin Liquid Lane review (06-2026, host not verified as auditor-controlled) lists a medium finding: . Its recorded status is partially fixed. A single bad feed can freeze share pricing, deposits and withdrawals of a Liquid Lane vault.
  • Fallback. No decentralized oracle network, TWAP or second source is described for these feeds. Incident history: no oracle-caused loss is recorded in the sources.
  • Assessment: Each Liquid Lane redemption asset is priced from one external feed. A hard on-chain min/max bound is cited only for the Midas feed, and none is cited for the Pareto oracle, which reads one virtual price. The sources do not name who updates the Midas feed. Before the Pareto fix, a defaulted asset was priced from the virtual price alone. After the partial fix, the defaulted token is valued at zero but intake stays enabled, which the audit flags as a remaining risk. The Midas weak point is a stale in-range price still passing. Whether the fixes are in the deployed contracts is not compared.

Flash Loan & MEV Exposure

  • No exploited flash-loan vector is recorded. The DeFiLlama hacks file for this family has no records .
  • Donation attacks (acknowledged or partly fixed). Bailsec rated "Vault composition can be overridden via donation attacks" as High. The status is partially fixed: the Morpho adapter tracks balances in storage, but "Other adapters however still use the samebalanceOf function" . In the audit's example, the attacker donates discounted shares of an illiquid underlying Morpho vault to the adapter and redeems Symbiotic vault shares for liquid assets from another adapter. The vault is left holding the illiquid Morpho shares . The audit notes the same attack was used on Morpho V1 vaults during a depeg.
  • Front-running of losses. Bailsec's Issue 05 is rated Low and acknowledged: . The text says slashable stake counts at full value until an authorized slash executes . Informed users can exit first; remaining holders absorb the loss.
  • Flash-loan control of adapter liquidity (Core V2 withdrawal queue). Bailsec describes a user taking a flash loan from Aave so that an adapter holds almost nothing, then calling sweepPending . Because of rounding, a redeem can send zero assets to the queue while burning a non-zero amount of shares, which needs a queue, a vault short of assets after bad debt or slashing, and constrained adapter liquidity . The recommended fix, reverting zero-asset redeems, was only acknowledged . The audit's example burns a specific share slice, and a related rounding case lets a claim advance for a dust slice without transferring assets . It is a griefing path with small illustrated losses, not a profit path for the attacker; the audit-to-deployed-version link is incomplete, so live exposure is unconfirmed.
  • No price floor on Liquid Lane conversions. Non-owner conversions use a delayed, permissionless path with no on-chain price floor, defended only by converter monitoring during a one-day delay . That is an operational defense, not guaranteed execution value (host not verified as auditor-controlled).
  • Sandwiching of swaps. Liquid Lane findings include "Discount swaps can be sandwiched for profit" (medium, acknowledged) . Rewards-related timing items (JIT deposits capturing rewards) are also acknowledged.
  • Mitigations in place. The Morpho adapter fix and a hard price bound on the Midas oracle. Against the front-running issue Bailsec wrote . The finding was only acknowledged, so no such delay is cited.

Tokenomics Risk: Yield, Leverage & Reflexivity

  • Yield sources (named, split not cited). Network rewards "may come from protocol fees, scheduled token emissions, external client payments, or amounts routed from slashing" . Liquid Lane lists three yield sources: redemption spread, base lending yield, and long-duration exposure . The split between revenue and token emissions is not published. The vault list shows low rewards figures (for example 0.11% to 4.28%) without a source breakdown . These are displayed figures, not audited facts.
  • Protocol fees. The Core V2 fee registry on-chain shows zero global management and performance fees at block 26296259 . Curators can set vault fees up to a 5% annual management fee and a 20% performance fee .
  • Points. Points are not a payout. Core V2 vault points accrue at a fixed rate on allocated assets . Points are not counted as paid yield.
  • Leverage. No on-chain LTV or leverage in the user's position is documented. Capital is lent to Morpho, Aave and Euler through adapters. Credit guarantees (Cap) use slashing to cover borrower defaults, so depositors take the credit risk: "the credit risk is transferred to the capital allocators providing the guarantees" . This is risk transfer, not looping, and no maximum exposure per borrower is documented.
  • Own token. The sources identify no Symbiotic governance or utility token backing or collateralizing vaults. SPK in the TVL composition (about 4.2%) is listed as a deposited token, not shown to be a Symbiotic token . No own-token reflexive loop is identified.

Solvency, Liquidation & Bad Debt

  • Backing and verifiability. Vault share value is read on-chain from vault state and adapter balances, so backing can be checked. Legacy DefaultCollateral wrappers report zero debt, but the audit explains that the debt variables are left unaltered and always equal 0 by design , so this zero does not indicate solvency or backing adequacy. For DC_wstETH, supply is 15,936 wstETH against a limit of about 420,729 wstETH, so the cap is not binding . The sources give no aggregate on-chain backing attestation for V2 vault shares as a whole.
  • Exit path and delay. V2 withdrawals use an instant path and a queue. Liquidity can be trapped in lending adapters. The docs say "Lending market utilization spikes can affect liquidity available for RWA redemptions and vault withdrawals." . They add that vault depositors may not get liquidity in time when the vault's delay is shorter than the asset's redemption window . No maximum delay is stated for Core V2 vaults. It is set per vault by the curator.
  • Loss handling. Slashed collateral goes to a per-vault Burner, whose policy (burn, redistribute or route) is set at deployment . In Liquid Lane V1, NAV impairment is not shared across depositors: "NAV impairment losses are not socialized (in V1 of Liquid Lane)" . The docs warn that this . No protocol-wide insurance or reserve fund was identified in the sources.
  • Slashing mode risk. The slasher type is fixed at deployment: instant execution or a short veto window . Instant slashers have no on-chain dispute window.
  • Liquidation. Symbiotic is not a liquidation-based lending market. Debt-like exposure sits in the applications (Cap) and the lending markets the adapters use. Liquidator access, incentives and bad-debt records for those systems are not in the sources. They are gaps.
  • Bad-debt record. No bad-debt report, risk-provider record or deficit account for the current version was found. Absence of a record is not proof of no bad debt. The DeFiLlama hacks file is empty (above).
  • Audit state of V2 accounting. Core V2 had Bailsec and Statemind reviews in 06-2026 . A Statemind finding that matured adapter withdrawal debt was not settled before a slash is marked fixed .

Incentive Alignment

  • Aligned parts. Slashing is bounded by a capture-time guarantee . Curators earn fees on the vault they manage, so they have an incentive to protect the vault. A bug bounty exists: Cantina, USD 500,000, for Core and Rewards .
  • Misaligned parts. Vault creation is permissionless, so any curator can open a vault . The curator picks adapters and risk, while depositors carry the loss. Slashing or adapter losses fall on remaining holders when informed holders exit first (Issue 05 above).
  • Mercenary capital. Points are tied to TVL allocated to adapters, which can attract short-term capital. The current share of TVL driven by points is not published. The 7-day DeFiLlama TVL change was -4.54% as of the 2026-10-29 snapshot, which is mild and not evidence of an exit wave .
  • Loss cover. No insurance, reserve, safety-module or compensation fund is identified. Docs describe insurance as a use case for third-party applications, not a Symbiotic backstop . No incident payouts are recorded (no family hacks).
  • Fund-control note. The 2-of-3 Safe at 0x5721ce64ee0d772ce613b62d411350091c544cd0 owns the factories and the DefaultCollateral limit setting, with no delay on setLimitIncreaser . It sets the V1 collateral caps. Governance is assessed separately.

Economic Attack Vectors

  • Price manipulation. The reachable inputs are Liquid Lane oracle feeds. Stale in-range prices from the Midas feed still pass, and the Pareto fix values a defaulted token at zero while intake stays enabled. These are audit-described risks. No exploit is recorded.
  • Donation or share-price attacks. Raw-balance accounting in non-Morpho adapters remains open in the audit (partially fixed). In the audit's example, the attacker gains liquid assets from the good adapter in exchange for donated illiquid Morpho vault shares, and the remaining holder is left with the illiquid position.
  • Slash-timing arbitrage. A holder who learns of a pending slash can exit through the instant path. Impact falls on remaining holders. Likelihood cannot be banded: no slash event or exit record is in the sources, and one would decide it.
  • Liquidation cascade and bad debt. Cascades in Morpho, Aave or Euler, or a Cap borrower default, would hit vaults through adapter losses or slashing. The sources contain no stress test or loss estimate. No fund to absorb it was identified in the sources.
  • Bridge risk. No bridge in the protocol itself. The deposit tokens (uniBTC, LBTC, and similar) are issued by other protocols and may rely on their own bridges or custodians. Those configurations are not in the sources.
  • Composability risk. Every strategy layer is a critical dependency for a depositor. Curators name these directly: . The risk page confirms this .
  • Further strategy layers. Vaults using ThreeFAdapter deploy assets into whitelisted 3F bridge facilitator requests , and the live principal stays illiquid until finalization . Liquid Lane conversions also depend on CoW Protocol orders. Live request exposures are not in the sources, and this audit's host is not verified as auditor-controlled.
  • Cap integration (source reliability: unverified publisher). Cap's own page states default borrower LTV of 50% and liquidation threshold of 80%, and that withdrawals take up to 14 days . These are unconfirmed, integration-specific claims, not measured parameters or leverage taken by Symbiotic depositors.
  • Dependency graph (code-owned). Backing tokens and their TVL shares: uniBTC 0.3038 , wstETH 0.2579, LBTC 0.147, rETH 0.0686, weETH 0.0608, SPK 0.0418, solvBTC 0.0286, osETH 0.0239, sfrxETH 0.0137, USDC 0.0103. All are unresolved tokens with no published score from us . No protocol-to-protocol edge carries a score or a usable_until date. The graph's warnings list is empty. No oracle provider edge is listed.

Aggregator Scores

  • DeFiLlama: USD 316.6M TVL, snapshot 2026-10-29, Ethereum only .
  • CertiK: a curated Skynet snapshot (captured 2026-10-24) gives 83.91, grade A, rating "good" . It is a point-in-time score and is not proof of current safety.
  • DeFiSafety and Exponential: no files available.
  • Comparison. CertiK's good rating is consistent with the heavy audit history. It does not measure the economic exposures above: curator discretion, adapter dependencies, oracle bounds and the lack of an identified loss-cover fund.

Data Completeness

  • Available: TVL and token composition, on-chain state at Ethereum block 26296259 (2026-10-29), mechanism documents (slashing, rewards, fees, Liquid Lane, credit), 2026 Core V2 audits and findings, CertiK score, empty hacks file.
  • Missing: top-10 depositor shares; revenue and emission split of paid yield; per-vault withdrawal bounds in Core V2; liquidator and bad-debt records for dependent lending markets and Cap; per-curator risk limits; DEX depth for rETH and sfrxETH wrappers and for V2 vault shares; the 0.5% depth band; any insurance or reserve fund; share of TVL tied to points.
  • Source caveats. Many 2026 audit files come from hosts not verified as auditor-controlled, per the provenance warning file. Audit-derived points keep lower confidence. The LlamaRisk research document (August 2024) is historical and was not used for current-state claims.
  • Confidence: moderate. The mechanism and audit evidence support the structure. The absence of loss-cover, yield-split and bad-debt data limits any stronger conclusion.

Score: 5.3/10

Rubric security-score-v2.5; data completeness 0.90

Economic security factors, points out of 10 (points)
Solvency redemption
5.50 points
Oracle design
5.50 points
Dependency risk
5.50 points
Liquidity concentration
3.50 points
Yield design
5.50 points
FactorLevelEvidenceWeightPoints
Solvency redemptionL2supported0.355.50
Oracle designL2supported0.255.50
Dependency riskL2supported0.155.50
Liquidity concentration—not disclosed0.103.50
Yield designL2supported0.155.50

74 citations · click markers to view evidence

👥

Team & Transparency

7.5/10

Social & Team Transparency: Symbiotic

Symbiotic is run by a publicly named team whose security and finance backgrounds are described on its own About page, and it publishes its contract addresses, many audit reports and lists a $500,000 bug bounty. No hack or exploit is recorded for it or its family in the DefiLlama hacks snapshot. The main gaps are the legal entity and jurisdiction, treasury and revenue reporting, and a complete outage and pause history. One structural point matters to users: the CEO and CTO co-founded Statemind, which is also one of Symbiotic's auditors. Independent verification of the founders' histories is missing, so confidence in the team findings is moderate; confidence in community metrics is low because no social-platform data was collected.

Team Identity

The team is doxxed: the official About page names three leaders with professional histories, but the operating legal entity and its jurisdiction could not be confirmed.

  • Misha Putiatin, Co-Founder and CEO. The page adds that he began as an auditor at KPMG and an equity research analyst at VTB Capital.
  • Algys Ievlev, Co-Founder and CTO. The same page says he was technical lead for Lido's expansion to Polkadot.
  • Jillian Friedman, COO. She also founded a crypto-focused law firm in 2014, according to the same page.
  • Staff hires are announced. The official newsroom lists .
  • Backers. The official Liquid Lane doc says . A LlamaRisk research page dated 2024-08-15 says . An OAK Research article (source reliability: unverified publisher; published 2024-09-24, updated 2025-12-05) gives both $5.6 million and $5.8 million for the round . A Series A of $29 million and a seed round "co-led by Paradigm and cyber•Fund" are reported by The Block (source reliability: unverified publisher) . Crunchbase (source reliability: unverified publisher) lists . These funding details rest on weak publications and are unconfirmed.
  • Legal entity. The homepage links a Terms of Service and a Privacy Policy . Two attempts to read the Terms PDF failed (timeout and invalid content), so the entity name is not confirmed. Crunchbase (source reliability: unverified publisher) gives a location of Grand Cayman, Cayman Islands , which is a headquarters location and not proof of incorporation. The jurisdiction is therefore not confirmed.
  • Independent verification of the founders' histories was not captured. Their profiles rest on the protocol's own About page.
  • Related-party point. Both the CEO and CTO co-founded Statemind. Statemind audit reports appear in Symbiotic's audit list, for example . Users should treat Statemind reviews as affiliate reviews, not fully independent ones.

Track Record

The family has operated since 2024-06-11 with no incident in the hacks dataset, and the contracts show continuous new deployments through 2026.

  • Operating history. Code-owned family age starts at the listing date 2024-06-11. The earliest Ethereum contract in the creation-date data is DefaultCollateral: rETH .
  • Delivery. The product moved from restaking vaults to "Collateral Markets" (credit, insurance, RWA). The official About page lists milestones including and . The current address list includes contracts created in 2026, such as LiquidLaneAdapterFactory, AppAdapterFactory and UniversalDelegatorFactory. No formal public roadmap was available, so roadmap adherence cannot be measured.
  • Prior projects of team members. The CEO and CTO co-founded Statemind (smart contract security), and the CTO also co-founded MixBytes, per the About page quotes above.
  • Incidents. The DefiLlama hacks snapshot holds no records for this protocol or its family . No post-mortem or rekt.news document was found in the collected sources, and a web search for hacks, pauses, outages or vault depegs from 2024 to October 2026 returned no report of such an event. That search does not establish a complete history. incidents is therefore empty, which means no qualifying incident was found, not proof that none occurred.
  • Own-token depeg history. The protocol has no measured own issued Ethereum token: the asset check reports . The assets list is empty.
  • Core contracts. The official address page lists VaultFactory and other factories, but individual vault proxies that hold deposits are not listed. No core_contracts row is given because the listed contracts are mostly factories and registries.
  • TVL context. DefiLlama reports about USD 316.6M on Ethereum only . This is a size context, not a misconduct signal.

Social Presence

Official channels are active, but quantitative community metrics were not collected.

  • GitHub. The org symbioticfi hosts many public repositories (core, relay, rewards, hooks, burners, collateral, network and others). The main branch of symbioticfi/core has commits in at least 9 of the 12 months from October 2025: October 2025, December 2025, January to June 2026 and August 2026 were each checked; July 2026 was not queried, and November 2025 shows none . The most recent commit to symbioticfi/relay-contracts is dated 2026-10-08 . Contributor counts were not captured.
  • Newsroom / blog. Official posts appear on 2026-07-27, 2026-08-19 and 2026-09-02 . That is roughly monthly in the captured window. Posts after 2026-09-02 were not captured. The blog also publishes technical explainers.
  • X/Twitter, Discord, Telegram. Links exist (a Telegram link appears on the official blog header). Follower counts, engagement and community quality: no data.
  • Third-party coverage. The About page links CoinDesk and The Block pieces and a Pantera article. Their text was not captured.

Transparency Indicators

Contract addresses, audit reports and bug-bounty terms are public; financial reporting is not.

  • Addresses. A first-party address page exists .
  • Audits. The official audit page lists many reports by Bailsec, Statemind, OpenZeppelin, Ackee, Cyfrin, Sigma Prime, Sherlock, Cantina, Certora, OtterSec, ChainSecurity and Zellic, with dates from 07-2024 to 08-2026. Examples: . Some reports are hosted only on the team's GitHub, so their provenance is unverified. That lowers provenance confidence only. Audit scope and findings are covered in the code-security report, not here.
  • Bug bounty. The official audit page lists the program: . A Cantina LinkedIn post (source reliability: unverified publisher) announces a USD 500,000 maximum reward . The program is listed but its live status on the Cantina platform page was not captured, so it is not confirmed as active.
  • Financials. No treasury, revenue or expense reports were found. No token or tokenomics information was found either.
  • Open source. Smart contracts are public on GitHub with source links from the address page. Licensing was not checked.
  • Incident disclosure. No incident or post-mortem was found in the collected sources.

Red Flags

  • Legal entity and jurisdiction not confirmed. This is a data gap, not evidence of anonymity.
  • Auditor affiliation. The CEO and CTO co-founded Statemind, a recurring Symbiotic auditor (see Team Identity). Other firms also audited the code, so the effect is limited.
  • No treasury or revenue disclosure found.
  • No evidence of social silence, deleted content, unrealistic promises or copied code in the sources.

Positive Signals

  • Three named leaders with security and finance histories described on the official About page.
  • Named institutional backers (Pantera, Paradigm) on the official docs.
  • More than two years of operation since the 2024-06-11 family listing, with no hacks in the DefiLlama snapshot.
  • Many published audits from multiple firms, and a USD 500,000 bounty listed on the official audit page (live status not confirmed).
  • First-party address list and open-source repositories.
  • Roughly monthly official newsroom posts in July–September 2026.

Data Completeness

  • Available: About page team section, audit and address pages, homepage newsroom, hacks snapshot, asset checks, Cantina LinkedIn post.
  • Also available: GitHub commit history for symbioticfi/core and symbioticfi/relay-contracts, a Crunchbase page and a The Block funding article (both unverified publishers).
  • Missing: no first-party or press articles in the article folder (zero included; one third-party research article, OAK Research, was captured as a web page); legal entity documents (the Terms PDF could not be read); contributor counts; Twitter/Discord/Telegram metrics; treasury and revenue reports; independent press confirmation of founders.
  • This analysis rests on collected docs and pages, not live social-media scraping.

Score: 7.5/10

Rubric security-score-v2.5; data completeness 1.00

Team transparency factors, points out of 10 (points)
Team identity
7.50 points
Disclosure
7.50 points
Development activity
7.50 points
FactorLevelEvidenceWeightPoints
Team identityL3supported0.457.50
DisclosureL3supported0.357.50
Development activityL3supported0.207.50

Track record: 8.0/10

Rubric security-score-v2.5; data completeness 0.80

Track record factors, points out of 10 (points)
Operating history
7.50 points
Incident record
10.00 points
Resilience
3.50 points
FactorLevelEvidenceWeightPoints
Operating history (computed)L3supported0.307.50
Incident record (computed)L4supported0.5010.00
Resilience—not disclosed0.203.50

29 citations · click markers to view evidence

🗳

Governance

2.8/10

Governance Analysis: Symbiotic

No voting process or governance token appears in the supplied sources. A small multisig, a 2-of-3 Safe with no timelock, controls the Core factories. The evidence supports a multisig-administered design; whether the current Core V2 contracts are upgradeable is not confirmed, because the immutability statements found date from 2024. The sources read do not show whether that multisig can upgrade contracts or move funds directly, so admin control is not classified; they do show it can change fees and approve new contract versions with zero delay. Users are mainly exposed to the owners of individual vaults and adapters, whose controls this review could not measure. Confidence is moderate: the Safe signers' identities and the vault-level admin setups are not disclosed.

Governance Structure

  • Type: multisig-administered protocol whose current Core V2 upgradeability is not confirmed. The supplied sources describe no DAO, no governance token and no voting venue, but none states outright that no vote exists.
  • Voting mechanism: none found. The code-measured vote file lists no venues . No Snapshot space or Governor address is named in the docs, so no voting venue is recorded.
  • Proposal process, quorum: no voting process is documented in the supplied sources, so how changes are decided is not disclosed.
  • Timelock: none on the Core factory and fee owner. On Ethereum at block 26296259 (2026-10-29), the ProtocolFeeRegistry fee setter shows . The same zero delay applies to factory whitelist(address) calls.
  • Emergency bypass: no separate emergency role appeared among the Core admin functions read. The owner Safe acts immediately for every listed function.
  • Vote concentration: not measurable, because no voting venue was found.

Admin Key Analysis

The Core factories are owned by a 2-of-3 Safe with undisclosed signers, and a second Safe holds the adapter registry. The sources do not establish whether the 2-of-3 Safe can upgrade contracts or move funds, and no timelock applies, so admin control is not assigned a level.

  • Factory owner. On Ethereum at block 26296259, the owner of VaultFactory (0xaeb6…a346, last version 3, 367 entities) and of the other Core factories is 0x5721…0544. That address is a Safe with threshold: 2 and owners_count: 3 . The three signer addresses are listed on-chain, but no source ties them to named people or teams.
  • Adapter registry owner. AdapterRegistry and AccountRegistry are owned by 0xf6bb…0b6a, a Safe with threshold 3 of 5 owners . Its signers are also unidentified.
  • Powers of the 2-of-3 Safe:
    • Whitelist or blacklist implementation versions on the factories. These calls show a zero delay .
    • Set global and per-vault fees and the fee receiver on ProtocolFeeRegistry. On-chain, the global management and performance fees currently read zero .
    • Raise collateral limits as limitIncreaser on the DefaultCollateral contracts (the rETH one is read).
    • Transfer or renounce ownership of each factory.
  • The docs describe the factory owner's role for the Liquid Lane adapters: The same audit adds that
  • Liquid Lane curator, converter and pauser powers. The Zeppelin Core v2 Adapters audit (29 June 2026) covers this audited design, and deployment of these adapters on-chain is not confirmed here. It says It also says
  • The Ethereum read is partial. Role-holder history is not complete for every contract, and 112 privilege rows are unresolved. Some migrate and curator-level functions could not be parsed.
  • Vault-level control. Vault creation is permissionless, so each vault owner or curator holds its own keys. The docs say curators set fees, deposit limits, adapters and allocation, and advise that That is guidance, not enforcement, and the vault owners of the 367 vaults are not assessed here.
  • Benefit: a small Safe can react fast, for example by blacklisting a bad implementation version. It also adds little decentralization. Fast reaction is limited, since the audit says blacklisting is only advisory on-chain.
  • Reasonableness: factory whitelisting and fee setting fit a factory owner. A 2-signature threshold with no delay is thin for powers that shape which code future vaults and adapters run.

Upgradability

The original 2024 Core is described as non-upgradeable, but the Core V2 and Liquid Lane audits describe a migratable pattern that the factory owner and entity owners control together, so current immutability is not established.

  • The LlamaRisk analysis (August 2024) states that It also says That description predates the current V2 deployment, so it is not confirmed for it.
  • The Chain Security audit of the Core contracts (13 August 2024) notes that It also notes . The same report says
  • Current components: VaultFactory reports lastVersion: 3. VaultFactory and UniversalDelegatorFactory expose migrate(address,uint64,bytes), whose guard the parser could not resolve. The Liquid Lane audit describes MigratableEntity as an upgradeable base:
  • The Bailsec Core V2 audit (June 2026) describes the factory as The core repository README says the vault owner migrates a vault to versions whitelisted by Symbiotic .
  • Assessment: a migration likely needs both the entity owner (vault or adapter curator) and a factory-whitelisted version, so the 2-of-3 Safe alone is unlikely to force a new implementation onto an existing entity. This rests on the README text and the unparsed migrate guard, and is not confirmed on-chain.
  • No timelock gates whitelisting or fee changes, as shown above.
  • Collateral tokens used by Symbiotic are third-party contracts. The on-chain read shows EIP-1967 admin slots set on swETH , and similarly on mETH and ETHx. Their admins are outside Symbiotic's control and are not assessed here. The upgrade guard of the USDC proxy was not resolved by the read.
  • Immutable versus upgradable list: because the available data do not give a complete per-contract list, only the factories and registries above are characterized.

Parameter Controls

  • Fees: the 2-of-3 Safe sets global and per-vault protocol fees on ProtocolFeeRegistry with no delay. Current global fees read 0. The Bailsec Core V2 audit says and the fee documentation caps vault fees at for management and for performance. These bounds come from audited code and documentation; the deployed implementation was not compared.
  • Collateral limits: increaseLimit and setLimitIncreaser on DefaultCollateral sit with the same Safe, again with zero delay.
  • Adapter whitelisting: setWhitelistedStatus on AdapterRegistry is held by the 3-of-5 Safe, with zero delay.
  • Vault parameters: curators set fees and limits per vault, and a 100% curator fee is possible in an earlier audit finding: . This is a vault-level risk for depositors and rewards, not a protocol-level admin power.
  • Harm potential: no direct fund-seizing path was found in the fee and limit functions read, though the full set of owner powers is not established. Fee changes on newly created or fee-charging vaults could reduce depositor returns without notice.

Decentralization Assessment

  • Position on the spectrum: centralized at the protocol-admin layer (a 2-of-3 Safe, no delay, unnamed signers), with some factory powers limited by fee caps and by the need for vault-owner action to migrate, though the deployed Core V2 immutability is not confirmed.
  • Roadmap: no documented plan to progressively decentralize admin control or move it to a DAO appeared in the sources.
  • Key-person risk: two of three signers on one Safe could change fees, limits and approved implementations at once. The sources do not name the signers, so independence between them cannot be judged.
  • Token distribution: no governance token is documented in the supplied sources, so voting power versus stake cannot be assessed.
  • Investors and team: the project description lists backers such as Paradigm and Pantera, but no source says they hold admin keys.

Governance Risks

  • Single points of failure: the 2-of-3 Safe controls factory ownership, fee setting and the rETH collateral limit. Loss of two keys allows instant, undelayed changes to whitelisting, fees and limits.
  • Compounding with curators: depositors in a vault also depend on that vault's owner. Admin setups of individual vaults are not measured. In the audited Liquid Lane design, the curator and the pauser can also affect redemption proceeds and redemption availability, as described under Admin Key Analysis.
  • Whitelisting risk: the audit says whitelisted implementations cannot be removed from the list, and blacklisting is advisory. A bad whitelist entry therefore has lasting reach.
  • Governance-attack vectors (flash-loan votes, bribery) cannot be assessed, since no voting venue was found. If none exists, there is also no community check on admin actions.
  • Centralization creep: the on-chain read shows the same Safe owning many Liquid Lane account and adapter factories, each with its own whitelist power. New integrations that reuse that owner would add to its reach.
  • Deployment default: an audit note says the default vault deployment script puts every vault role on one address . Production vaults may differ, and this review did not measure them.

Data Completeness

  • Available: on-chain owner, threshold and delay data for the Ethereum Core factories at block 26296259 (2026-10-29); a LlamaRisk description of non-upgradeable core (August 2024); audit text on factory whitelisting and migratable entities.
  • Missing:
    • Signer identities of both Safes. A targeted web search found no source naming the signers or linking them to the Symbiotic team.
    • Complete role-holder histories. The on-chain read hit its time budget on 19 addresses and an address cap on others, so it is partial.
    • Per-vault owner configurations across the 367 VaultFactory entities.
    • A statement of whether any governance process exists. A targeted web search found no governance token, DAO, Snapshot space, Governor or timelock in the pages it captured.
    • Whether the current V2 Core is non-upgradeable.
  • Not assessed: other chains (only Ethereum was read) and the third-party collateral token admins.

Score: 2.8/10

Rubric security-score-v2.5; data completeness 0.30

Governance factors, points out of 10 (points)
Admin control
3.50 points
Change delay
1.00 points
Decision process
3.50 points
FactorLevelEvidenceWeightPoints
Admin control—not disclosed0.453.50
Change delayL0supported0.301.00
Decision process—not disclosed0.253.50

26 citations · click markers to view evidence

🔗

Cross-Analysis

Cross-Risk Analysis: Symbiotic

Symbiotic holders carry three risks that no single report shows in full. First, withdrawals from Core V2 vaults depend on adapter and oracle layers that auditors left partly open. Second, about 90% of TVL sits in BTC wrappers and ETH liquid staking tokens, with no identified loss-cover fund. Third, a 2-of-3 Safe with no delay controls the factories, while the deployed VaultFactory differs from its audited revision. No hack is recorded, so these are structural exposures, not realised losses. Confidence is moderate: the audit-to-deployment link, per-vault admin setups and per-vault exposure to each adapter are missing.

Top Risks for a Holder

1. Core V2 vault withdrawals can freeze or leak value through the adapter and oracle layer (Core V2 vault shares, Liquid Lane vaults, Morpho/Aave/Euler-adapter vaults).

  • What goes wrong: instant exit depends on idle assets and instantly-deallocatable adapter assets. Open audit items can stall or distort that path. One frozen redemption account can block deallocation and all vault withdrawals (acknowledged, not fixed). A reverting oracle feed can brick accounting (partially fixed). Donation attacks on non-Morpho adapters stay open (partially fixed). Economic analysis adds that lending-utilisation spikes can reduce the liquidity available for redemptions.
  • Magnitude: a freeze, not a certain loss. Duration is unbounded in the sources, because the maximum delay is set per vault by the curator. Funds at risk are the vaults using these adapters. The share of the USD 316.6M TVL in them is not given. One cited single vault holds "Tvl 67.45M". A donation or timing attack shifts value from remaining holders to the attacker. In the audit's example the vault is left holding illiquid Morpho shares.
  • Likelihood: medium for a temporary freeze in at least one vault; low for a profitable exploit. No exploit is recorded, but most of these items were only acknowledged. The audits describe reviewed commits and were not compared with deployed code. Many sources are hosted on unverified hosts, which lowers confidence, not the score.
  • Missing fact that would change the band: which vaults use Liquid Lane or non-Morpho adapters, and whether deployed code includes the fixes.

2. Concentrated collateral in BTC wrappers and ETH liquid staking tokens, with no loss-cover fund (uniBTC, LBTC, wstETH, rETH, weETH vault deposits).

  • What goes wrong: a depeg, custody failure or bridge failure at one issuer hits every vault holding that token. Symbiotic vaults inherit the issuer risk, and no reserve absorbs it.
  • Magnitude: uniBTC USD 96.29M plus LBTC USD 46.61M is about USD 143M, roughly 45% of TVL. wstETH, rETH, weETH, osETH and sfrxETH add about 42% by share, from the dependency graph. DEX depth is measured only for the legacy DC_wstETH wrapper, which showed zero 2% depth; depth for the listed collateral tokens themselves is not measured. That wrapper redeems 1:1 for its backing asset, so the gap is in market exit, not in redemption.
  • Likelihood: low. Basis: it needs an external failure at an issuer, custodian or bridge, which Symbiotic does not control; no such impairment is recorded in the sources, and no issuer failure rate is available. The sources do not contain issuer-level controls that would raise or lower this band. The tokens are unresolved in the dependency graph with no score from us. The issuer configurations (custody, bridge) are not in the sources. Impact if it happens is severe because the exposure is correlated. Route to the dependency-risk owner for reassessment.

3. A 2-of-3 Safe with zero delay and unnamed signers controls factory versions, while the deployed factory code differs from its audited revision (all Core vaults; 367 VaultFactory entities).

  • What goes wrong: the Safe can whitelist or blacklist implementation versions and set fees with a measured fastest delay of 0 seconds. Two signers can approve a bad version at once. The 2024 Core audit says a whitelisted implementation cannot be removed later. The advisory-blacklisting finding is stated for the Liquid Lane adapter and account factories, so it is not applied to VaultFactory here. Whether the 2024 removal limit still holds for the current VaultFactory is not confirmed. The deployed VaultFactory differs from the audited version in 4 files, and the code check judged the change material (it matches the audited OtterSec commit symbioticfi/core@788f2d7 as the comparison base). The Bailsec Vaults V2 review excluded admin-key risk. The reports do not show any review of the Safe's powers against the factory code.
  • Magnitude: the governance review found no direct fund-seizing path in the fee and limit functions, and a migration likely also needs the vault owner. The realistic harm is a malicious version reaching new vaults or vaults whose owners migrate. Funds at risk would be the deposits of those vaults. The sources give no count of vaults that migrate or use affected versions, so the amount is not quantified; that count would size this risk.
  • Likelihood: low. Basis: a two-signer quorum is needed; one signer is labelled "Symbiotic: Deployer", and none is a named person; there is no recorded abuse since the family listing in 2024-06-11. That record mostly covers the earlier Core: the docs date Core V2's launch to July 2026. The missing facts are the signers' identities, the guard on migrate and whether the 2024 removal limit applies to the current VaultFactory, any of which would move the band.

Combined Attack Vectors

Vector 1: Slash-timing exit with no loss cover

  1. A credit or insurance application (for example Cap) holds a slashable guarantee over a vault. A borrower default or claim becomes likely.

  2. The vault still counts slashable stake at full value until an authorised slash executes, so an informed holder withdraws on the instant path. The Bailsec finding was only acknowledged, and no withdrawal delay is cited.

  3. The slash lands on the remaining holders. The slashing documentation describes slasher modes and a capture-time cap, but it belongs to the earlier-generation docs, and the docs read do not show how Core V2's AppAdapter maps onto them. This vector therefore relies on neither the slasher mode nor the cap.

  4. No reserve or insurance absorbs the loss.

  5. Separately, after a loss a user could use an Aave flash loan to constrain adapter liquidity and call queue processing. Rounding can then burn queue shares without delivering assets. The economic report calls this a griefing path with small illustrated losses. These are sequential conditions, not a demonstrated combined exploit.

  • Domains: mainly economic design (exit race, no loss cover), with the acknowledged audit finding on loss timing and the structure around it: the credit application sits on top of the vault, and no global pause or shutdown was found. This is close to a single-domain scenario. Assessment: no documented control halts the instant path once a loss becomes likely, because the sources show no global pause; per-vault curator controls were not measured.
  • Separate item, not part of this sequence: a Statemind medium finding on slashing being bricked by an unbounded claimable backlog is marked fixed. It does not feed the exit race above.
  • Likelihood: medium. The exit race needs only a foreseeable default or claim, which a credit application with live borrowers can produce; no slash event is in the sources, so the frequency of such losses is the missing fact. Informed insiders can trigger the exit part without any exploit.
  • Impact: moderate to severe for the affected vault, minor for the protocol as a whole.

Vector 2: Liquid Lane price or feed failure becomes a vault-wide freeze, with curator power as amplifier

  1. A redemption asset's feed goes stale in range, reverts, or the asset defaults. The Midas oracle still accepts stale in-range prices. The Pareto fix values a defaulted token at zero but leaves intake enabled.
  2. Share pricing, deposits and withdrawals of the vault can freeze (single-feed failure, partially fixed). Dust balances below a redemption minimum can freeze sync and conversion.
  3. An attacker can sandwich discount swaps and permissionless CoW conversions, which have no slippage protection or on-chain price floor.
  4. A curator-registered converter can sell proceeds at any price, which the audit calls the most powerful trusted capability, and a pauser can deny service.
  • Separate, version-limited risk (not part of this sequence): the docs state that "NAV impairment losses are not socialized (in V1 of Liquid Lane)", which may create withdrawal races. The sources do not establish that this V1 rule applies to the audited V2 adapter, so it is not used above.

  • Domains: oracle design + code (OpenZeppelin and Bailsec each report overlapping issues; the OpenZeppelin report sits on a host not verified as auditor-controlled) + governance (curator and pauser powers).

  • Likelihood: medium for a freeze, low for deliberate theft. Basis: the failure inputs are external issuers' feeds, and the abusive powers sit with curators who are chosen per vault. Liquid Lane audit sources are on unverified hosts.

  • Impact: moderate. Scope is limited to Liquid Lane vaults; how much TVL they hold is not stated. The governance report notes that deployment of the audited adapters on-chain is not confirmed.

Vector 3: Curator handover leaves stale converter authority

  1. A curator transfers ownership of a Liquid Lane adapter, and holders assume the new owner controls every redemption account.
  2. Existing accounts keep stale owner and converter authority (acknowledged, not checked against code).
  3. A retained converter then uses the broad conversion power to sell proceeds at an adverse price.
  • Domains: account-lifecycle code + curator authority and key rotation.
  • Likelihood: low. It needs an affected implementation, a handover and misuse of retained authority. Live account permissions would change the band.
  • Impact: severe for the proceeds held by the affected accounts; the amount is unknown.

Vector 4: Safe compromise leads to a malicious vault version

  1. An attacker compromises two of the three Safe keys. The signers are unnamed, so key hygiene and independence cannot be judged.
  2. The Safe whitelists a malicious implementation with zero delay. The 2024 Core audit says whitelisted implementations cannot be removed; whether that holds for the current VaultFactory is not confirmed, and the advisory-blacklisting finding is stated only for the Liquid Lane factories.
  3. New vaults deployed from it, or existing vaults migrated by owners who trust the whitelist, run the hostile code. The default deployment script gives all vault roles to one address, which would make a vault owner a single point; production vault setups were not measured and may differ.
  4. Detection depends on community monitoring, because there is no timelock, no voting venue and no global pause.
  • Domains: governance (zero delay, thin quorum) + code (deployed factory differs from audit; the 2024 audit says whitelisted versions cannot be removed) + social (signers unattributed).
  • Likelihood: low. It needs two keys and a second step by vault owners.
  • Impact: severe for affected vaults, but bounded by the second step.

Vector 5: Relay validator-set or key-registration flaw (separate component from Core vaults)

  1. The code check found missing BLS subgroup checks still present in the latest relay-contracts code, and three high items still appear. The Statemind report that raised them is AI-generated.
  2. Three other critical items were not located or not checked, including unauthenticated key setters.
  3. A Sherlock finding says a malicious operator could control consensus without risking stake after unstaking. The team acknowledged it.
  4. A network that relies on Relay attestations (for example a bridge or appchain) could accept forged attestations while slashing does not reach the operator.
  • Domains: code + economic security (stake that backs attestations) + social (Statemind affiliation, so the AI report carries limited independent weight).
  • Likelihood: low to medium, with the band unresolved. The deployed Settlement contract has no attributed address, so whether and where this code is live is not known. The latest-repository code may differ from what is deployed.
  • Impact: severe for any relying network, minor for Core vault depositors (no shared funds are shown).

Systemic Risk Factors

  • External token dependencies (route to the dependency-risk owner for reassessment). The dependency graph (composition measured 2026-10-29) lists 10 backing tokens: uniBTC 30.38%, wstETH 25.79%, LBTC 14.7%, rETH 6.86%, weETH 6.08%, SPK 4.18%, solvBTC 2.86%, osETH 2.39%, sfrxETH 1.37%, USDC 1.03%. All are unresolved, with no published score and no usable_until. The graph's warnings list is empty, so no circular backing or downstream holder is flagged.
  • Strategy layers. Every layer a deposit passes through (Morpho, Aave, Euler, Pareto, Midas, Centrifuge, CoW, 3F) is a dependency for the depositor. This belongs to the dependency-risk owner.
  • More strategy layers. ThreeFAdapter vaults commit principal to 3F bridge-facilitator requests until finalisation, and Liquid Lane conversions depend on CoW Protocol orders. The Cap and Nexus Mutual applications sit on top of vaults. These are supported paths, not measured allocations.
  • Graph limits. The empty warnings list means no circular-backing or downstream-holder warning is recorded; it does not prove their absence.
  • Chain. Ethereum only, so no sequencer risk, though Ethereum congestion would hit exits and monitoring together. Bridge risk enters through the wrapped tokens (uniBTC, LBTC), whose bridge configuration is not in the sources. It also enters through the 3F bridge-facilitator requests that ThreeFAdapter vaults use; live 3F allocations are not measured.
  • Correlation. BTC wrappers (about 48% by share, summing uniBTC, LBTC and solvBTC) and ETH liquid staking tokens (about 42%, summing the five tokens) move together in a market shock. A lending-utilisation spike at Morpho or Aave would hit the adapter exit path at the same time (see Top Risk 1).
  • Macro. A stablecoin depeg touches mainly Liquid Lane (USDC vaults for redemptions); USDC itself is 1.03% of TVL.

Risk Amplifiers

  • Unlinked audits plus fast-changing code. Only 1 of 25 deployed rows is shown identical to an audited version, and the audited findings are marked acknowledged for most medium items. A fix may not be live, and an acknowledged item may be live.
  • Affiliated reviewer on the V2 core. The CEO and CTO co-founded Statemind. Of the two direct Vault V2 audits cited in the overview, only Bailsec is independent, and it excluded admin-key risk. The overview counts two Vault V2 audits; the independent count is one.
  • Zero delay plus no global pause plus no voting. Factory-level changes have no waiting period, no global pause or shutdown was found, and no voting venue gives a community check. Liquid Lane differs: a pauser can pause all swaps, and non-owner conversions have a one-day execution delay.
  • Curator discretion plus permissionless vault creation. Anyone can create a vault, and curators pick adapters. Risk then depends on vault choice, which a newcomer cannot judge from the listing.
  • Points-linked deposits plus a mild -4.54% 7-day TVL change. Points accrue on assets allocated to adapters, which can attract short-term capital. The share of TVL tied to points is not published.
  • Acknowledged-only findings plus unverified hosts. 20 of 35 audit reports sit on GitHub hosts not verified as auditor-controlled. That lowers confidence in the status of those findings but is not a penalty.

Risk Mitigators

  • Breadth of review. 35 audit reports from about 12 firms. Bailsec and OpenZeppelin each report overlapping Liquid Lane issues. The OpenZeppelin review is hosted where auditor control is not verified, so the overlap shows attributed coverage, not independent confirmation.
  • Formal verification and competition on the earlier Core (Certora, Cantina). These do not cover Core V2.
  • Bug bounty. Cantina, USD 500,000, for Core and Rewards, listed on the official page; live status and whether V2 and Liquid Lane are in scope are not confirmed.
  • Bounded slashing (earlier-generation docs only). The earlier-generation slashing documentation says a slash cannot exceed the guarantee captured earlier. The docs read do not show how Core V2's AppAdapter maps onto it, so this is not counted as a mitigator for Core V2 vaults.
  • Fee caps and zero protocol fee. Vault fees are capped at about 5% management and 20% performance, and global fees read 0 on-chain at block 26296259.
  • Legacy wrappers redeem 1:1 for their underlying token, so thin DEX depth is not a trap for wrapper holders.
  • Track record. Two-plus years without a recorded hack in the DefiLlama snapshot, active development, a doxxed leadership team. The record is absence of evidence, not a clean-history proof.
  • Split registry control. A separate 3-of-5 Safe holds AdapterRegistry, so adapter whitelisting does not rest on the same 2-of-3 Safe. No signer of either Safe is tied to a named person, and the two Safes share no listed owner address.
  • Core V2 fixes. Statemind's Core V2 review marks all 5 medium findings fixed. It is an affiliate review and does not resolve the Liquid Lane findings or show deployed fixes.
  • Reactor is identical to its audited commit, the only fully matched row.

Contradictions & Inconsistencies

  • "Core vault accounting is audited" versus audit-to-deployment gaps. The economic report says the core vault accounting is audited. The code report shows VaultV2 has no attributed deployed address and 177 further audit names are not compared. The audited claim holds for the reviewed commit only. The economic wording should be read as "audited at a commit", not "audited as deployed".
  • Overview counts two direct Vault V2 audits; the independence view counts one. Statemind is affiliated (social, code), so the independent coverage of Vault V2 is Bailsec alone.
  • CertiK 83.91 (grade A) versus code-security 4.3 and 4 open critical items. The sources reconcile this only by noting that CertiK is a point-in-time snapshot that does not measure the economic exposures. The open critical items concern Relay, a different component from the Core vaults CertiK's score would reflect; the sources do not state what CertiK's score covers.
  • Incident record versus unresolved findings. The social report scores the incident record at the top level (no hacks) while the code report scores unresolved findings at zero. Both can hold: no exploit, but a large, mostly unfixed finding list. The first is evidence of past outcomes, not of safety.
  • Governance "admin control not assigned a level" versus the overview's "low to moderate" abuse likelihood. The governance low score (2.8) is driven by missing data (a cap for unknown admin control and low data completeness 0.30), not by a measured abuse path. This analysis keeps likelihood low for the same reason.
  • "Non-upgradeable" (LlamaRisk, 2024) versus migratable V2. The overview cites the 2024 immutability statement; governance reports a migratable pattern in V2. Governance treats current immutability as unconfirmed, which is the safer reading.
  • Cap ">$250m collateral" versus TVL of USD 316.6M. The sources do not say whether the Cap figure is inside the DefiLlama TVL, so the share of TVL under credit-slashing risk cannot be computed.
  • Bounty certainty. The overview states the bounty covers Core and Rewards; the social report says its live status was not captured. The narrower reading is kept: a listed program, not verified cover.
  • Oracle identification. The overview says the Liquid Lane price source is not identified; the economic report identifies Midas feeds and a Pareto virtual price but not who updates them. The mechanism is partly identified.
  • Set-aside theft claim. An OpenZeppelin finding that in-flight swap balances can be stolen by arbitrary callers was set aside by the code check, so this report does not use it as an attack path.
  • Operating history versus V2 history. No recorded hack since 2024 does not show that Core V2, launched in July 2026, has been stress-tested.
  • Governance says no fund-seizing path in fee and limit functions; the same report says a Liquid Lane curator converter can sell proceeds at any price. These are different layers (factory owner versus vault/adapter curator), so they do not conflict, but "no seizing path" applies only to the factory owner.

Overall Risk Profile

Symbiotic is a tier-2 collateral-markets protocol with about USD 316.6M of TVL on Ethereum, a doxxed team, many audits and no recorded loss. Its composite risk comes from layering: depositor funds pass through a curator's vault, adapters, external lending or redemption protocols and issuer-controlled tokens, and no global pause, no factory-level delay and no loss-cover fund were identified. The most likely failure mode is a temporary withdrawal freeze or a value transfer from remaining holders to informed exiters in a single curator's vault, caused by an adapter, oracle or slashing event (Vectors 1 and 2). The worst case is a correlated issuer failure in uniBTC or LBTC (about USD 143M, 45% of TVL) with no identified reserve and unmeasured DEX depth for those tokens, or a malicious factory version approved by two unnamed keys with no delay (Vector 4). The inputs show strengths in team disclosure and audit volume, and weaknesses in factory-level delay, audit-to-deployment traceability and vault-level transparency. The sources contain no peer data, so no ranking against category peers is made.

Data Gaps

  • Collection of GitHub repositories stopped at the collection limit; some material may be missing.
  • Current deployment addresses could not be confirmed for some audited contracts.
  • Source summaries cover only the first 80000 characters of 11 document(s); later content was not assessed.
  • Audit-to-deployment link. 77.8% of compared lines have no identified audited version; VaultV2, Settlement and 177 other audit-named components have no attributed address. Vector 1 to 5 live exposure cannot be confirmed without it.
  • Per-vault exposure. TVL by vault and adapter type (Liquid Lane, Morpho, Aave, Euler, 3F), holder concentration and the share driven by points are not published. Top Risk 1 cannot be sized.
  • Cap and credit exposure. Per-borrower limits, the overlap of Cap's collateral with TVL, and any stress test are missing. Vector 1 cannot be sized.
  • Signers and migrate guard. Safe signer identities and who may call migrate are unresolved; 112 privilege rows are unresolved. Vector 4 depends on both.
  • Issuer risk of the collateral tokens. Custody, bridge and redemption terms for uniBTC and LBTC are not in the sources. Top Risk 2 depends on them.
  • Relay deployment. Which networks rely on Relay and whether the audited code is live is unknown.
  • Rewards findings. Replay and unclaimable-reward findings sit in the code report, but pending rewards and live contract linkage are unknown, so no principal-loss vector is built on them.
  • Own token. No Symbiotic governance or utility token is identified, and deposited SPK is not shown to be Symbiotic-issued, so no own-token loop is assessed.
  • Unchecked findings. 81 of 135 selected findings were not checked against code for lack of check budget, so the open medium items in Liquid Lane and Vaults V2 rest on audit status alone.
  • Dex depth. Only DC_wstETH was probed; DC_rETH and DC_sfrxETH errored, and V2 vault shares were not measured.

80 citations · click markers to view evidence

Preview3 · Default / Rows