Author: @LlamaRisk
Date: 2025-04-14
Summary
It’s been a year since our last renewal. LlamaRisk (LLR) submits this proposal to renew our role as Aave’s Risk Service Provider.
Aave’s risk layer previously depended on infrastructure the protocol did not own, could not inspect, and could not operate independently. The recent transition away from one of the protocol’s risk providers highlighted this dependency, leaving Aave without its automated risk-oracle infrastructure, its parameter automation pipeline, and its Risk Steward. Recognizing the urgency of the situation, LlamaRisk stepped up, assumed control of the Risk Steward together with @AaveLabs (AL), and immediately began executing a plan to ensure operational safety and continuity across all Aave markets. The immediate question is continuity. The structural question is how to ensure the protocol is never in this position again.
The answer is protocol-owned risk infrastructure. When risk management is delegated to a single external provider operating proprietary, closed-source systems, the protocol inherits a dependency it cannot inspect, verify, or replace without disruption. This is the challenge Aave is now working to address. LlamaRisk’s proposal resolves this structural vulnerability by building Aave’s risk layer on Chainlink’s Runtime Environment (CRE), where the Aave DAO retains full control of the code and can independently verify off-chain logic through cryptographic workflow IDs.
What we will deliver: LLR will absorb all departing risk functions and ensure full risk coverage across V3, V4, and Horizon through five pillars:
- Protocol-owned risk infrastructure on CRE. Risk-managed price feeds (Pendle PT, CAPO, USDe, RWA NAV, etc.), dynamic parameter automations (supply/borrow caps, interest rates, Umbrella calibration, credit lines, risk premiums, liquidation configuration, etc.), and safety mechanisms (USDe freeze guardian, V4 per-spoke circuit breakers, etc.). All code audited with @Certora, all deployments co-owned with AL.
- Full ownership and accountability of the risk function. Risk Steward co-ownership, oracle design in close collaboration with Chainlink, AIP/governance payload review, and Guardian signer duties. LLR assumes direct operational liability for parameter updates across all markets, not as a delegated advisor but as an accountable partner.
- Unified communications and institutional presentation. Coordinated communication with AL, other service providers , and external partners. Shared monitoring, tooling, and dashboards that give Aave DAO’s key stakeholders real-time visibility. Private escalation protocols and standardized playbooks that present a coherent and unified image to stakeholders, preserving and upholding the Aave brand at a time when institutional credibility matters most.
- Risk-native infrastructure and R&D. Custom V4 spokes (liquidator spoke for RWAs, instant settlement bridge), and applied research designed to create durable competitive moats for Aave. Collaborative development with shared ownership: LLR builds, Aave DAO owns.
- Doubling down on RWA. Horizon co-operation, LlamaGuard NAV expansion, and the Horizon Vision 2026 roadmap, all without allowing Aave to become an exit venue for brittle RWAs. As V4’s Credit Lines unify liquidity, Horizon evolves from a separate Hub into a deeply integrated part of Aave’s architecture.
How we are structured. LLR is a team of 16 professionals, scaling to 20+ over this engagement, entirely self-governed and free from external investor mandates. We propose to phase out the remaining non-Aave scope over six months, going all-in on Aave and sending a clear signal to the Aave ecosystem and its partners that LLR’s commitment is unconditional. By the midpoint of this engagement, LLR will be an Aave-exclusive organization. We fully align with AL’s leadership role, commit to private coordination on all sensitive matters, and will work in close collaboration with Chainlink on Oracles and decentralized infrastructure, @Certora on security and auditing, and with @TokenLogic on growth, incentives, and treasury operations.
What it costs to get the job done. $4M for one year, structured as $1.5M upfront and $2.5M streamed linearly. The upfront component funds the immediate absorption of all risk-related scope, hiring, and infrastructure buildout required for an uninterrupted transition. The fee reflects expanded scope and the structural costs any risk provider at this scale must absorb: cybersecurity infrastructure proportional to tens of billions in protocol deposits, and the institutional overhead of consolidating into an Aave-exclusive organization.
Motivation
1. Our Role Evolved: From Watchdog to Core Infrastructure
It has been two years since LLR began serving the Aave Protocol. We entered as a lean, qualitative risk supplement: a second opinion on asset onboardings, a fresh pair of eyes on parameter decisions.
That is no longer what we do. Over three (3) epochs, our role has expanded across every dimension of the Aave Protocol’s risk function. On V3, we deliver a full suite of risk frameworks, parametrization methodologies, and quantitative models across all active deployments. We serve as Guardian signers. This is the operational backbone of the protocol’s risk coverage, not a supplementary opinion.
On V4, we have produced exhaustive research and frameworks covering the full feature set: Hubs & Spokes risk architecture, the Reinvestment Controller, Umbrella coverage logic, credit line risk limits, Liquidation Engine parameter modeling, and cross-pool contagion modeling. This is embedded, collaborative R&D conducted over many months in close coordination with AL and SPs, translating directly into the launch strategy and risk architecture that will define V4.
On Horizon, we jointly operate the protocol with a comprehensive due diligence, parameterization, and risk-monitoring framework, having pioneered innovations such as the bounded dynamic NAV model (LlamaGuard), which enable safe RWA integration at scale. We build automated risk infrastructure on Chainlink CRE, with LlamaGuard evolving into protocol-owned infrastructure that extends Aave’s risk layer to offchain compute. The on-chain Parameter Registry makes all off-chain computation values publicly verifiable; LlamaRisk has no privileged access controls; and every parameter update is cryptographically verifiable via CRE workflow IDs. This empowers Aave rather than exposing it to the pitfalls of opaque third-party dependencies.
Across all three verticals, we serve as the Aave Protocol’s only independent legal and regulatory research capability and as the only independent check on delegated risk authority. The scope has consistently outgrown the compensation, and the role must now expand further.
2. What Has Changed
The sudden departure of Aave’s primary risk provider has created a structural concentration of operational risk that must be addressed, not just through personnel replacement. Two recent incidents illustrate why, and a critical gap in Oracle coverage shows where the opportunity lies.
2a. Structural Vulnerability Exposed
On March 10, the wstETH CAPO oracle failed, resulting in approximately $1.03M in borrower damages, 47 wrongful liquidations, and over 4 hours of depressed pricing. Weeks earlier, our analysis (“Retro: WETH utilization spike and Slope2 Risk Oracle performance”) documented a significant divergence in the Slope2 risk oracle’s behavior during a utilization spike, revealing gaps in methodology that had not been disclosed.
These are evidence of a structural gap: concentrated operational authority with no independent oversight. The code was proprietary, the methodology was not disclosed to AL or other SPs, and there was no independent verification layer. LLR had no mechanism to detect, flag, or block either failure.
2b. Yield-Bearing Assets and the Oracle Gap
Aave DAO has expressed a clear interest in accelerating the onboarding of yield-bearing stablecoins. Our USDe risk oracle and freeze guardian proposal (October 2025) directly addresses this: a multi-source pricing decision tree combined with a circuit breaker that freezes all USDe-priced reserves in response to stress signals. The Resolv incident (March 2026, $25M exploit draining Morpho and Fluid via hardcoded $1 valuations) validates exactly this threat model against Aave’s $6.4B+ exposure. This framework generalizes to other yield-bearing assets (such as sUSDai and beyond), and its absence today compounds the structural risk described above.
3. What We Are Proposing
Protocol-owned risk infrastructure is the centerpiece: every element below connects to the thesis that Aave’s risk layer must be owned by the Aave DAO, not rented from a third party. We will work in close collaboration with Chainlink to design the workflows, with @Certora on security auditing of all CRE deployments, and with @TokenLogic on growth initiatives, incentive design, and treasury operations to ensure risk parameters support rather than hinder protocol growth.
3a. The LlamaGuard Suite: Risk Oracle Deployment Roadmap
Building on our track record in risk frameworks, quantitative modeling, and CRE, including the successful implementation of dynamic bounded NAV feeds for Aave Horizon, the next period will focus on deploying a comprehensive suite of risk oracles for the Aave Protocol.
We are ready to take over the Manual Risk Steward and will immediately work towards progressive CRE automation, with fully audited code (in collaboration with @Certora), tackling the highest-priority integrations first. Aave DAO retains full control: co-ownership of the CRE admin panel, cryptographic verification of running code via workflow IDs, and no update or shutdown without Aave DAO’s explicit consent.
Immediate continuity: absorbing departing scope
LLR has developed quantitative and heuristics-based methodologies for all V3 and V4 parameters, and is already capable of fully operating changes using the Manual Risk Steward. We will progressively migrate each function to CRE automation, prioritizing PT token oracles as the highest-urgency integration.
| # | Integration | Impact |
|---|---|---|
| 1 | Pendle PT Price Oracles on CRE | Fully on-chain methodology, provably lower overpricing risk for $2B+ PT exposure; deployed via Chainlink CRE |
| 2 | Aave V3 and V4 Interest Rate Analysis | Full IRM coverage: Slope1, Slope2, Base rate, Uopt; tie-ins with growth, direct protocol revenue impact |
| 3 | Dynamic Supply & Borrow Caps | Revenue impact, high-visibility across 12+ chains |
| 4 | Aave V3 and V4 SVR Monitoring & Configuration | Risk parameters informed by SVR auction data pipeline, searcher competition analysis, MEV recapture optimization |
| 5 | Aave Umbrella | We authored the ratified methodology; coverage calibration anchored to external risk factors, extending to cross-Spoke contagion awareness in V4 |
| 6 | CAPO Risk Oracle | Foundational LST/LRT/stablecoin pricing; CRE-based pre-execution validation |
| 7 | USDe Oracle + Freeze Guardian | Highest-impact new integration: $6.4B+ Ethena exposure |
| 8 | Dynamic RWA NAV feeds | LlamaGuard NAV is already operational |
New capabilities: V4-native parameter spaces
Entirely new parameter categories introduced by V4. No incumbent. LLR has already produced the foundational research for each.
| # | Integration | Description |
|---|---|---|
| 9 | Credit Line & Draw Caps | Hub & Spoke centerpiece; concentration risk, cross-Spoke correlation |
| 10 | Aave V4 Liquidation Analysis & Configuration | Target HF calibration, variable liquidation bonus curves, SVR feedback loop |
| 11 | Freeze Agents (V4) | Per-Spoke circuit breakers; generalized from Ethena design |
| 12 | Dynamic Risk Premiums | Per-user interest surcharge based on collateral quality |
| 13 | Reinvestment Controller Limits | Optimize idle capital deployment, protect withdrawal guarantees |
Each integration is detailed in our internal planning document, with sequencing recommendations ready for joint prioritization with AL and SPs.
3b. V3: Continued Full-Stack Coverage
V3 will remain critical infrastructure for at least the next year as V4 migration proceeds. Our V3 scope includes all existing risk frameworks and parametrization, monitoring, AIP payload reviews, PT rollovers, oracle design in close collaboration with Chainlink, Umbrella calibration, caps management, interest rate analysis, SVR monitoring, and governance participation. This is not maintenance; it is active risk management across 12+ chains during a complex migration period.
3c. V4: Scope We Are Already Building
V4’s Hub & Spoke architecture, credit lines, isolation pools, and expanded delegation mechanisms create risk surfaces that must be parameterized from inception. LLR has not waited for V4 to launch. We have produced exhaustive research and frameworks covering the full V4 feature set, in active collaboration with AL and other SPs.
Our Horizon work is the proving ground: every model built for Horizon, from bounded dynamic NAV to dynamic liquidation bonus frameworks, is forward-compatible with V4’s architecture. LlamaGuard is architected as protocol-owned infrastructure, where the Aave DAO retains full control and visibility across the entire stack, extending Aave’s risk layer to offchain compute and enabling risk-managed pricing strategies for asset types (equities, long-duration credit, commodities, custodial accounts) that are otherwise impossible to safely parameterize.
3d. The Reinvestment Controller
The Reinvestment Controller enables V4 to deploy idle capital into productive strategies while preserving supplier withdrawal guarantees. Our role optimizes allocation caps based on evolving protocol-side conditions, in coordination with @TokenLogic on strategy selection, deployment, and incentive design.
The Horizon Vision 2026 roadmap is phased across three stages: Phase I (dynamic liquidation bonus models, pricing infrastructure for tokenized equities, LlamaGuard CRE migration), Phase II (Hub & Spoke migration, spoke isolation, permissionless RWA strategies, GHO RWA issuance spoke), Phase III (RWA borrowability, risk tranching via Umbrella, advanced liquidation solutions). We intend to continuously evolve this roadmap jointly with AL and SPs; this should be an iterative process, with strategy closely aligned.
3e. The Aave Foundation and LLR’s Role as Board Member
LLR is actively interested in serving as a board member of the newly created Aave Foundation. LLR is the only service provider combining deep protocol knowledge with legal and regulatory expertise. Our CLO leads dedicated legal research covering stablecoin regulation (MiCA, GENIUS Act), custody and counterparty risk, DeFi lending compliance, tokenized RWA legal-risk cartography, and institutional legal structuring. We will continue working closely with AL and SPs on legal and policy matters.
3f. Collaborative R&D
The model is collaborative development with shared ownership: LLR develops MVPs in close coordination with AL and other SPs (joint concept validation, joint audit oversight with @Certora), and the resulting infrastructure is protocol-owned from day one. This is the model already working for LlamaGuard: LLR builds, Aave DAO owns.
R&D tracks include: the liquidator spoke (essential for RWAs and self-custodied integrations lacking secondary-market liquidity, and a prerequisite for scaling Horizon beyond assets with robust secondary markets) and the instant settlement bridge (aligned with the vision to internalize looping infrastructure).
3g. Improving Risk Models and Tooling
Continuous investment in the infrastructure that underpins day-to-day risk operations:
- Simulation and modeling. Expanded stress-testing frameworks for V4-specific risk surfaces (cross-spoke contagion, credit line utilization dynamics). Dedicated RWA simulation engine for settlement friction, redemption buffer depletion, and NAV deviation scenarios.
- Dashboards and monitoring. Real-time risk metrics across all Aave deployments with Horizon-specific modules. An independent observation layer that tracks every parameter update across all markets and feeds them into the proposed veto window. Expanded SVR monitoring dashboard.
- Horizon-specific tooling. LlamaGuard NAV on CRE with expanded asset type coverage. Dynamic parameter management suite: time-weighted dynamic liquidation bonuses, off-hour protection modules for tokenized equities, and corporate event handling automation.
- War room and incident response. Formalized practices following the model built for Aave Horizon. Updated SOPs for incident escalation. Automated alerting systems feed into the private escalation channel with AL and SPs.
4. The Fee: Why $4M Is Fair
The $4M unified fee for a 1-year renewal covers V3, V4, and Horizon under a single agreement, replacing the previous split structure. Under the “Aave Will Win Framework,” all protocol revenues flow to the Aave Protocol. A single agreement reflects this reality.
Fee Structure: $1.5M Upfront + $2.5M Streamed Linearly
The upfront payment funds immediate absorption of all departing scope, team scaling, and infrastructure buildout. It also signals mutual commitment by allowing LLR to phase out all non-Aave engagements over a 6-month period. The remaining $2.5M is streamed linearly over 12 months, aligning compensation with delivery.
Responsibility and Structural Costs
Co-signing authority means direct operational liability for parameter updates across all markets. Foundation board membership introduces governance accountability. This is a qualitative shift from advisory to operational. The fee also accounts for typically invisible costs: cybersecurity infrastructure proportional to tens of billions in deposits, key management and operational security, and the institutional overhead of consolidating into an Aave-exclusive organization. The fundamental asymmetry of DeFi risk management is that upside is incremental while downside is uncapped. The $4M fee makes this asymmetry manageable.
Team Scaling: 16 Today, 20+ by Year-End
We plan to onboard at least four new full-time contributors. Our needs may evolve as we expand Aave v4 and validate custom Spokes development. We are building a plan to maximize in-house capability without relying on contractors; every new hire will be embedded in the Aave scope from day one.
| Hire | Role | Primary Assignment |
|---|---|---|
| 1 | Backend Engineer | CRE risk oracle deployment pipeline, improved monitoring infrastructure |
| 2 | Quantitative Researcher | V4 parameter modeling, credit line calibration, risk premium methodology |
| 3 | Smart Contract Developer | Risk Steward integration, governance payload review, Spoke development |
| 4 | Team Lead (Risk Oracles) | Lead CRE risk oracle engineering, deployment coordination, and infrastructure scaling |
5. Private Escalation and Relationship Maturity
LLR commits to private escalation to AL and SPs before any public communication on sensitive matters. This supports the unified front necessary to accelerate convergence with TradFi and build trust with retail and institutional counterparties. The power of the Aave brand depends on this discipline, and LLR is structured for it. There is no place for politicians in the next phase of Aave’s growth. The protocol must present an institutional-grade front to convince TradFi counterparties and neobanks.
6. Closing Statement and Our Commitment
This proposal reflects what we’ve built together over the past two years- and, more importantly, what we are ready to build next. The relationship between LlamaRisk and Aave DAO is not a vendor arrangement; it is a partnership anchored in shared infrastructure development, aligned incentives, and a common ambition to make Aave the most resilient protocol in DeFi.
It’s easy to underestimate how risk compounds quietly- oversight gaps widen, parameter integrity weakens, and decisions lose rigor. Confidence erodes, and when stress hits, the system reacts rather than leads. Sadly, such casual attitudes have undermined the efforts of many talented builders in our space.
But Aave is built differently. At its core, it has a deep appreciation for iron-clad security and rigorous risk management.
The $4M fee, structured with $1.5M upfront, is our commitment to embody those values. It enables us to fully align with Aave’s roadmap, scale the team beyond 20 dedicated contributors, and deliver the protocol-owned risk infrastructure that strengthens resilience while reducing long-term dependency. Our goal is not just to support Aave- but to help it operate with the level of certainty, independence, and robustness that its position in the ecosystem demands.
Specification
Proceed with the immediate payment of 1.5m GHO, and create a payment stream of 2.5m GHO to address 0x9eE16dBDE572886342fc1e2Db8525DEFB007b27c a LlamaRisk-controlled multisig for 1 year. Terminate current GHO stream (ID = 100071).
Next Steps
- Gather community feedback on this ARFC.
- If consensus is reached, escalate this proposal to the ARFC snapshot stage.
- If the ARFC snapshot outcome is YAE, escalate to the AIP stage.
Copyright
Copyright and related rights waived via CC0.