[ARFC] Add rsETH to Aave V3 Ethereum

[ACI]

Title: [ARFC] Add rsETH to Aave V3 Ethereum

Author: ACI (Aave Chan Initiative)

Date: 2024-08-12

---

ARFC has been updated with Risk Parameters provided by both Risk Service Providers after having reservations with exchange rate oracles. 2024-08-12

Summary

Kelp DAO is seeking community support for adding its Liquid Restaking Token, rsETH, to Aave V3 on Ethereum. In addition, rsETH depositors into Aave will accumulate additional Kelp miles and EigenLayer points.

Motivation

KelpDAO (Kelp Liquid restaking | restake ETH, stETH & ETHx) is one of the largest liquid restaking protocol built on top of the Eigen Layer. Restakers on Kelp get access to multiple benefits like restaking rewards, staking rewards and DeFi yields.

Latest stats:

• USD 800 Mn+ (~270k Eth) TVL
• 20K+ restakers onboarded
• Accepts ETHx, stETH, sfrxETH, and Native ETH as collateral
• Live across mainnet and L2s such as Arbitrum and Polygon zkEVM

KelpDAO has undergone multiple security audits by top tier audit firms including Sigma Prime, Code4rena. More about KelpDAO’s security here: Kelp DAO Audit reports.

The motivation for this proposal is to introduce a new asset class LRT (Liquid restaking tokens) to Aave and add to the diversity of LST/ LRT category.

Restaking Innovation

KelpDAO’s innovation in the restaking ecosystem has been focused on bringing about the following benefits to users and services-

• Restakers
  • Liqudity for restaked assets (stETH, ETHx, sfrxETH and Native ETH)
  • Access to additional staking rewards
  • Participate in DeFi using rsETH
• AVSs
  • Bootstrap economic security
  • Lowered incentive spends
• Operators
  • Higher revenue earning potential
  • Service onboarding and community support
  • Marketing for node delegation

Specification

> Risk Parameters have been provided by Risk Service Providers on 2024-08-09 and ARFC has been updated. It is recommended listing rsETH using an exchange rate oracle.

Parameter	Value
Isolation Mode	No
Borrowable	Yes
Collateral Enabled	Yes
Supply Cap	19,000
Borrow Cap	1,900
Debt Ceiling	-
LTV	72%
LT	75%
Liquidation Penalty	7.5%
Liquidation Protocol Fee	10.00%
Variable Base	0.0%
Variable Slope1	7.00%
Variable Slope2	300.00%
Uoptimal	45.00%
Reserve Factor	15.00%
Stable Borrowing	Disabled
Flashloanable	Yes
Siloed Borrowing	No
Borrowable in Isolation	No
E-Mode Category	ETH-Correlated

CAPO

maxYearlyRatioGrowthPercent	ratioReferenceTime	MINIMUM_SNAPSHOT_DELAY
9.83%	monthly	14 days

Detailed explanation on Kelp’s technical architecture - Technical Architecture - Kelp

KelpDAO will have its own token $Kelp that will be the main governance token of Kelp protocol. Kelp token holders will be able to participate in governance related matters including Whitelisting/ delisting of Assets, Node operators, AVS’, setting and updating fees for Kelp protocol.

Kelp’s on-chain stats can be viewed on Dune.

Liquidity Pools for rsETH

Following are liquidity stats for rsETH:

• rsETH has >55 Mn USD liquidity across several DEXs on the mainnet including
  • Balancer - $38 Mn
  • Uniswap V3 - $12.42 Mn
  • Balancer - $5 Mn
  • Few others
• ~$10 Mn+ average daily trading volume (Last 7 days on 12th May)

Oracles:

rsETH is supported by Chainlink and Redstone oracles. Posted links below to the Oracle contracts:

Chainlink calculated price feed: Kelp DAO: Chainlink Price Oracle | Address 0x78C12ccE8346B936117655Dd3D70a2501Fd3d6e6 | Etherscan

Chainlink market feed: RSETH / ETH | Chainlink

Redstone market feed:

https://app.redstone.finance/tokens#/app/token/rsETH\\ETH

rsETH exchange rate mechanism:

rsETH exchange rate is defined based on the underlying assets and rewards generated by those assets.

Accepted assets include ETH, stETH, ETHx, sfrxETH.

Exchange rate calculation -

A → Total ETH + Staking rewards accumulated from the chain

B → [Amount of ETHx] * [Value of 1 ETHx in ETH]

C → [Amount of sfrxETH] * [Value of 1 sfrxETH in ETH]

D → [Amount of stETH] * [Value of 1 stETH in ETH]

E → Total supply of rsETH

Exchange rate (rsETH / ETH) = [A + B + C + D] / E

Withdrawals and risks

Prior to this ARFC, the only way to exit rsETH was via DEXs.

Nevertheless, KelpDAO users can now unstake rsETH, with an unbonding period of 7 to 8 days, due to Eigenlayer withdrawal period. That also means that any of the LST’s including stETH, ETHx or sfrxETH from Eigenlayer can be unstaked as well.

Proof of Liquidity and Deposit Commitments:

Anyone who deposits rsETH into Aave will accumulate additional Kelp miles and EigenLayer points.

Useful Links:

• KelpDAO’s Dapp
• Contract Address: 0xA1290d69c65A6Fe4DF752f95823fae25cB99e5A7
• GitHub Docs

Disclaimer:

This proposal is powered by Skywards. The Aave Chan Initiative is not directly affiliated with Kelp DAO and did not receive compensation for creation this proposal.

Next Steps

1. Publication of a standard ARFC, collect community & service providers feedback before escalating proposal to ARFC snapshot stage
2. If the ARFC snapshot outcome is YAE, publish an AIP vote for final confirmation and enforcement of the proposal

Copyright:

Copyright and related rights waived under CC0

[ChaosLabs]

Overview

Chaos Labs supports listing rsETH on Aave V3 Ethereum as part of the community’s strategy to increase the offering of Aave protocol with more LST/LRT assets. Following is our analysis and risk parameter recommendations for the initial listing.

Note: The following analysis is conducted solely from a market risk viewpoint, excluding centralization and third-party risk considerations. If the community aims to reduce exposure to rsETH, adopting more conservative supply and borrow caps should be considered.

Liquidity and Market Cap

The amount of ETH restaked with Kelp DAO has grown rapidly, from 65K ETH in mid-January to 255K ETH today. Note that withdrawals were opened on May 4, 2024, which led to a 10K ETH decrease in TVL.

Currently, users may stake using ETHx, WETH, stETH, or sfrxETH. Rewards are reflected in the price of the token rather than rebasing. Since inception, it has registered $6.67M in average daily volume.

On Ethereum, its liquidity is concentrated on Balancer V2 ($38M TVL, paired with WETH) and Uniswap V3 ($14M TVL, paired with WETH).

rsETH / ETH Volatility

Liquidation Threshold

Considering the volatility and the correlation of rsETH to ETH, we recommend setting the LT to 75%, mirroring weETH on Ethereum V3.

Supply Cap, Borrow Cap, and Liquidation Bonus

Chaos Labs’ approach to initial supply caps is generally proposed through setting the Supply Cap at 2x the liquidity available under the Liquidation Penalty (7.5%) price impact, which leads us to recommend a starting supply cap of 8,000 rsETH.

Based on our observations, the utilization rate for LRTs and LSTs has been low due to additional borrowing costs. Therefore, we have taken a conservative approach and recommend a borrow cap of 800 rsETH. However, if there is a significant increase in demand and utilization, perhaps due to external incentives, we will reassess the caps according to the utilization pattern.

IR Curve Parameters

We recommend aligning the interest rate parameters with those of weETH on Ethereum to ensure consistency across similar assets on Aave v3.

Oracle Setting and E-mode

Following analysis, we find that rsETH has consistently traded below the reward-adjusted ETH price implied by its exchange rate. This persistent discount has been punctuated by periods of a widening discount, reaching nearly 3% in March and breaching 2% in April. The limited history since withdrawals were enabled shows that the discount remains, albeit with less severe spikes.

Given these factors, we first recommend utilizing the market price when listing this asset and not a calculated price feed. Additionally, we do not recommend listing this asset in E-Mode, given its tendency to face wide discounts, which could spark liquidation cascades in a highly leveraged market using a market oracle price. Below are our updated specifications.

Recommendations

Following the above analysis, we recommend listing rsETH with the following parameter settings:

Parameter	Value
Isolation Mode	No
Borrowable	Yes
Collateral Enabled	Yes
Supply Cap (rsETH)	8,000
Borrow Cap (rsETH)	800
Debt Ceiling	-
LTV	72.50%
LT	75.00%
Liquidation Bonus	7.50%
Liquidation Protocol Fee	10.00%
Variable Base	0.0%
Variable Slope1	7.00%
Variable Slope2	300.00%
Uoptimal	45.00%
Reserve Factor	15.00%
Stable Borrowing	Disabled
Flashloanable	Yes
Siloed Borrowing	No
Borrowed in Isolation	No
E-Mode Category	N/A

CAPO recommendations:

maxYearlyRatioGrowthPercent	ratioReferenceTime	MINIMUM_SNAPSHOT_DELAY
8.75%	monthly	7 days

[MarcZeller]

Before the escalation of this proposal, I would like to request an Oracle evaluation about the rsETH pricing model.

I’m yet to be convinced an Index approach is safe; we currently have large growth and “infinite” demand with weETH, and we thus have the luxury to be conservative with other LRTs even if diversity is appreciated.

[zmanian]

Hello Aave community, Zaki, cofounder of Sommelier

Sommelier is working with Kelp on a deeper vault integration with Kelp that would make extensive use of Aave looping through weth and steth on Aave.

As Sommelier powered vaults have continued to drive a significant amount of utilization on Aave and we would expect there to be a substantial amount of adoption driven by the vault.

[ChaosLabs]

After further analysis of the rsETH price, we have updated our recommendation in the original post above with a recommendation to exclude the asset from E-mode should the community decide on listing the asset.

[LlamaRisk]

LlamaRisk is in favour of @ChaosLabs’s recommendation for cautious onboarding of Kelp’s rsETH, including exclusion from e-Mode (as opposed to weETH and osETH), a limited supply cap of 8,000 rsETH, and a conservative LT of 75%. This recommendation stems from our initial review of rsETH, with a particular focus on its internal exchange rate (probed by @MarcZeller), and associated dependancies.

We share our preliminary findings below, emphasizing using rsETH market price for its integration with Aave. We are working on a comprehensive assessment of rsETH and will update the Aave community accordingly.

rsETH Exchange Rate Mechanism

The rsETH exchange rate is determined by the underlying assets (ETH, stETH, ETHx, sfrxETH) and the accumulated staking rewards. The LRTOracle contract stores the Oracle contract for each asset.

• stETH (oracle contract): price is hardcoded to 1:1 with ETH.
• ETHx (oracle contract): price relies on the getExchangeRate function of the StaderStakePoolsManager contract, which Stader updates via a permissioned function. The oracle does not independently check for price anomalies or liveness, and it is a proxy contract that the Stader team can upgrade. The getExchangeRate function, operated by Stader, returns a state variable called exchangeRate, updated by the submitExchangeRateData function, callable only by whitelisted addresses controlled by Stader. The ETHx price is determined by the “book value” of ETHx, calculated as totalAssets / totalSupply, similar to the ERC-4626 accounting method.
• sfrxETH (oracle contract): price is obtained from Frax’s sfrxETH pricePerShare function, an ERC-4626 tokenized vault. The price feed assumes frxETH to be 1:1 with ETH.

Market Price vs. rsETH exchange rate

The rsETH exchange rate calculation includes several assumptions, resulting in a considerable spread between the computed exchange rate and the market price of rsETH. This can be seen in the chart below (data from March 8th May 31st):

Source: rsETH internal rate, Redstone & Chainlink feeds and Uniswap V3 rsETH / ETH Pool (blocks 19390000 to 19987000)

Risks with dependencies

The main risk factors associated with rsETH’s dependencies are:

1. sfrxETH/stETH oracle assumptions: If the market price of stETH or frxETH falls below the ETH price, the collateral backing rsETH could be worth less than expected by the rsETH system. This could lead to a depreciation in the market price of rsETH.
2. ETHx oracle centralization: The ETHx price is obtained from a relatively centralized source. If a malicious actor manipulates the price feed, they could unjustly inflate the rsETH supply up to the ETHx deposit limit.
3. LRTManager permissioned function: The Kelp External Admin (6/8) is whitelisted as the LRTManager, which can updatePriceOracleFor without a timelock. This centralization could lead to mistakes or malicious actions that could brick the protocol or cause a loss of funds. For example, adding a new supported token requires calling three different functions in a specific sequence. If the calls are contained in other transactions, the protocol could be in an intermediate state, blocking the updateRSETHPrice() function.

[LlamaRisk]

Collateral Risk Assessment - Kelp DAO Restaked ETH (rsETH)

LlamaRisk conducted a comprehensive collateral risk assessment on rsETH, which can be read here. Our findings are generally favorable, with the Kelp team being receptive to our feedback and implementing several beneficial changes. These include displaying the 10% ETH staking reward fee and underlying breakdown on the UI and committing to implementing a bug bounty program.

We recommend onboarding rsETH with the earlier parameters: exclusion from e-Mode (as opposed to weETH and osETH), a limited supply cap of 8,000 rsETH, and a conservative LT of 75%. rsETH liquidity has increased an estimated 42% across different venues since our first assessment in late May 2024, and we perceive its risk as lesser now.

Below are the highlights of our research, supporting our recommendation. This recommendation is contingent upon implementing a bug bounty program, which we deem crucial and has been scheduled to go live in the coming days.

Risk Management summary

This section will summarize the report’s findings by highlighting the most significant risk factors in the three categories: Market Risk, Technology Risk, and Counterparty Risk.

6.1.1 Market Risk

LIQUIDITY: Does the LSD have a liquid market that can facilitate liquidations in all foreseeable market events?

The on-chain liquidity for rsETH is relatively good, with multiple integrations and liquidity venues. The minted rsETH is also relatively spread out among wallets and protocols, with no single EOA concentrating a significant portion of the supply. Top holders include LayerZero, a Zircuit re-staking pool, and a Pendle rsETH fixed-rate market. Liquidity pools with rsETH are either paired with WETH, ETH, ETHx from Stader, or weETH from EtherFi.

VOLATILITY: Has the LSD had any significant depeg event?

The rsETH/ETH secondary market rate has consistently traded at a small discount compared to the protocol’s internal exchange rate. However, the peg has recently improved due to the addition of ETH withdrawals and a reduced withdrawal delay from 7 to 2 days. The only significant depeg event was a -1.5% deviation in late April, which quickly corrected. Notably, as rsETH is partially backed by LSDs, its exchange rate with ETH would be affected by any LSD depeg. This is an inherent characteristic of an LRT backed by LSDs, not a fault of Kelp. By using the internal exchange rates of LSDs, KelpDAO provides rsETH/ETH with a more stable exchange rate.

6.1.2 Technology Risk

SMART CONTRACTS: Does the analysis of the audits and development activity suggest any cause for concern?

KelpDAO has conducted three audits on its codebase from renowned auditors. Although no serious flaws were revealed, some fixes were made. The codebase is public on GitHub, well-documented, and shows professional development practices. However, online development activity is limited to a few commits; no PRs or tagged releases are visible. A Code4rena audit competition allowed anyone to disclose vulnerabilities, but the total reward was limited to $28k, which is low compared to industry standards. Kelp has stated that a bug bounty program will be implemented shortly, which we consider crucial before onboarding.

DEPENDENCIES: Does the analysis of dependencies (e.g., oracles) suggest any cause for concern?

KelpDAO directly depends on the internal exchange rates of ETHx and uses a hardcoded exchange rate of 1 for stETH. While beneficial for exchange rate stability, this could enable an arbitrage attack if one of the supported LSDs were to depeg. The KelpDAO team has provided a mitigation strategy, but it remains imperfect and could still result in a net loss for KelpDAO users in some cases. The implementation of a circuit breaker should improve this issue. The rsETH/ETH exchange rate, used for deposits and withdrawals, relies on the price of its underlying assets.

KelpDAO also relies on off-chain services for some aspects of the protocol. These services operate through an EOA whose private keys are stored in an AWS secret manager, authenticated using a role-based access control system. Although no funds can be stolen if the private keys leak, a more robust and secure approach is warranted. A decentralized set of off-chain services with an on-chain threshold consensus could enhance reliability and security.

6.1.3 Counterparty Risk

CENTRALIZATION: Are there any significant centralization vectors that could rug users?

KelpDAO remains very centralized, with the development team having significant power over contract upgrades, protocol parameters, and total control over the off-chain services needed for minting and withdrawals to work correctly. There is no DAO nor governance tokens. Communication mediums for users exist, including Twitter, Telegram, and Discord, but they mostly serve marketing and support purposes. However, a role-based access control system and two different multisigs, along with a timelock with a 10-day delay for contract upgrades, provide a solid foundation upon which more decentralization should be built.

LEGAL: Does the legal analysis of the Protocol suggest any cause for concern?

Kelp’s Terms of Service require users to self-acknowledge compliance with sanctions and ensure their funds are not linked to illegal activities. However, Kelp does not enforce access restrictions at the user interface level. Users must confirm they are not sanctioned or from sanctioned regions and agree not to use Kelp for unlawful purposes. Kelp may conduct “Know Your Customer” and “Anti-Money Laundering” checks and terminate services if users provide false information.

Advertised as a non-custodial protocol—even though the protocol is relatively centralized with the team having control over it — KelpDAO claims that users are fully responsible for their digital assets and wallet security, with Kelp bearing no liability for third-party services or issues arising from wallet use. While the terms of services are clear, the Panama incorporation results in a lack of regulatory clarity.

Risk Rating

The following chart summarizes a risk rating for rsETH as collateral based on the risks identified for each category. The rating for each category is ranked from excellent, good, ok, and poor.

• We rank rsETH good in liquidity for the many liquidity venues that provide it. The existence of DEX pools paired with ETHx, based on the team’s relationship with Stader, is a plus.
• We rank rsETH good in volatility because the quality of its peg has increased consistently. The only depeg event was short-lived and limited in strength.
• We rank rsETH ok in smart contracts because of the limited number of issues in the three publicly available audit reports. The code source is public, well-documented, and professional. However, the lack of a bounty is a significant issue, given the importance of TVL.
• We rank rsETH poor in dependencies because of using a fixed exchange rate for stETH and internal exchange rates for ETHx, potentially socializing specific asset risks between all depositors. The lack of transparency regarding off-chain services is also a cause for concern. Although essential to the correct operation of the protocol, they are operated in a centralized way through EOAs.
• We rank rsETH ok in decentralization because the team still has significant control over the protocol, and there is no clear path to decentralization. Validators are trusted by three professional node operators, which is good but could be greater. The role-based access control system, the multisigs, and the Timelock are positive.
• We rank rsETH good in legal for establishing a fully compliant legal structure in the Panama Islands. Although user terms are clear, there still needs to be regulatory clarity from Panama’s securities laws.

Source: LlamaRisk comparative chart

Although KelpDAO has taken positive strides in making rsETH a suitable collateral asset, we generally advise caution when assuming exposure to any liquid restaking token. The sector remains highly speculative, driven by points programs that may lead to rapid shifts in demand for the asset, potentially resulting in sustained depeg events. LRTs are generally a less mature asset class with substantial centralization vectors and reliance on off-chain services that often involve processes overseen by the protocol team.

We advise that KelpDAO must address concerns critical to its security and transparency before being considered for collateral onboarding. We have been working with the KelpDAO team to resolve what we consider blockers. Their team has been receptive and has committed to addressing these points.

• Implementing a bug bounty program is urgently needed given the protocol’s high TVL, and any delay poses significant risks.
• KelpDAO’s communication needs improvement, particularly regarding changes like the unannounced removal of sfrxETH.

In addition to immediate blockers, we await the transition to a fully decentralized DAO structure, which should be expedited to ensure proper governance and community involvement and reduce reliance on off-chain services. Progressive decentralization is a common theme for many LRTs and DeFi protocols generally, but the current centralized management level highlights the protocol’s immature state.

After addressing the two primary points described above, we advise a conservative approach to onboarding with low exposure limits, conservative LT parameters (exclusion from e-mode in Aave, for instance), and a gradual onboarding process that matches the protocol’s maturity over time.

[Agvs]

Thanks Llama team for the assessment on Kelp’s rsETH. Addressing the immediate concerns below:

1. Bug Bounty program launch is in the final stages of review by legal and expected to be launched in a few days. Will update this with the actual program once it goes live.

2. Changes related to sfrxETH have already been posted here informing the community: Update for sfrxETH deposits on Kelp DAO | by Kelp DAO | Jul, 2024 | Medium

Will have a detailed response on the others soon.

[DheerajBorra]

Hi! Dheeraj from Kelp DAO here.

Bug bounty program for Kelp DAO is live on Immunefi - Kelp DAO Bug Bounties | Immunefi | Immunefi

[ChaosLabs]

Following the changes mentioned above, Chaos Labs recommends listing rsETH using an exchange rate oracle. We provide updated parameters, including a CAPO recommendation, below.

Analysis

At the time of our previous recommendation on May 16, withdrawals had just begun, and rsETH still faced a wide discount relative to its reward-adjusted ETH price. Because of this, we recommended using a market rate oracle, as well as excluding the asset from E-Mode.

However, ongoing withdrawals have proved beneficial to rsETH’s peg, decreasing from just under 100bps at the time of writing to 35bps now. Despite a temporary depeg during the August 4 crash, it has proved to mean revert quickly.

There were previous concerns regarding rsETH’s exchange rate model, which uses index pricing based upon the underlying assets. Following the removal of sfrxETH from rsETH’s backing, we note that the derivative assets underlying rsETH’s exchange rate — ETHx and stETH — are both included on Aave V3 using an exchange rate oracle. Given this, we find limited additional risk associated with listing rsETH using an exchange rate oracle, noting that this will allow the protocol to avoid liquidation cascades during temporary secondary market depegs.

Regarding the recommended supply cap, we have utilized our usual supply cap methodology, setting it at two times the liquidity available beneath the Liquidation Penalty. While rsETH’s supply did decrease following withdrawals opening, we have observed a leveling off in recent weeks.

Additionally, on-chain supply would have to fall to 25.3K rsETH — from 250K rsETH now — to violate the community-agreed rule in which supply caps do not exceed 75% of on-chain supply.

Specification

rsETH

Parameter	Value
Isolation Mode	No
Borrowable	Yes
Collateral Enabled	Yes
Supply Cap	19,000
Borrow Cap	1,900
Debt Ceiling	-
LTV	72%
LT	75%
Liquidation Penalty	7.5%
Liquidation Protocol Fee	10.00%
Variable Base	0.0%
Variable Slope1	7.00%
Variable Slope2	300.00%
Uoptimal	45.00%
Reserve Factor	15.00%
Stable Borrowing	Disabled
Flashloanable	Yes
Siloed Borrowing	No
Borrowable in Isolation	No
E-Mode Category	ETH-Correlated

CAPO

maxYearlyRatioGrowthPercent	ratioReferenceTime	MINIMUM_SNAPSHOT_DELAY
9.83%	monthly	14 days

[LlamaRisk]

We endorse @ChaosLabs’ parameter recommendations, including the ETH-correlated e-mode and the use of an internal exchange rate. We note that there have been no substantial qualitative changes since our last review in late July, except for the implementation of a bug bounty program, which we had strongly advocated for.

[ACI]

After new feedback from Risk Service Providers, the current proposal has been escalated to ARFC Snapshot. It has been included latest risk parameters.

Vote will start tomorrow, we encourage everyone to participate.

[ACI]

In parallel to ARFC Snapshot being opened for vote, an extra analysis from BGD Labs will be provided, in order to be aligned with Adding rsETH to Aave V3 Ethereum.

[ACI]

After Snapshot monitoring, the current ARFC Snapshot recently ended, reaching out both Quorum and YAE as winning option, with 892K votes.

Therefore, the ARFC to Add rsETH to Aave V3 Ethereum has passed.

Nevertheless, as stated in previous comment, an extra analysis from BGD Labs will be provided, before moving the proposal forward.