AVS-risk methodology

This section defines how Lyrasing reasons about LRT collateral before any protocol implementation exists. The goal is not to publish a finished oracle or parameter engine. The goal is to make the inputs, failure modes, and policy defaults explicit enough that a future implementation can be reviewed against a builder-grade AVS-risk methodology.

Price inputs remain necessary. A money market still needs market price, liquidity, volatility, redemption depth, and liquidation path evidence. They are not sufficient for LRT collateral. An LRT can trade close to ETH while carrying different operator allocations, AVS opt-ins, slash conditions, withdrawal windows, and insurance coupling. Lyrasing therefore treats price as one input inside the risk method, not as the method itself.

Methodology surface

The methodology separates five surfaces:

Surface	What the section owns
Observable inputs	What can be observed directly, what depends on primary protocol disclosures, and what to do when inputs are stale or missing.
Exposure taxonomy	How to classify direct AVS exposure, correlated operator exposure, shared vault or collateral paths, withdrawal coupling, and insurance coupling.
Slashability and timing	How capture timestamps, active guarantee windows, withdrawals, veto or appeal phases, and settlement flows affect risk before liquidation can react.
LTV adjustments	How observable risk evidence should compress caps, delay admission, or reduce LTV without pretending a numeric oracle is already implemented.

The glossary terms AVS exposure, slashing, collateral factor, insurance pool, and oracle/risk methodology are the anchors for this section.

Policy posture

The default posture is conservative because Lyrasing is pre-launch and this repo does not contain on-chain code, oracle contracts, supported collateral configuration, live data feeds, or a working app. When a required input is missing, the methodology should prefer one of four outcomes:

Missing or weak evidence	Conservative response
The LRT has price data but opaque AVS allocation.	Delay admission or assign only a research placeholder, not a collateral factor.
Operator set or operator overlap cannot be observed.	Compress any cap and treat correlation as unresolved.
Slash conditions exist only in off-chain prose or discretionary process.	Require primary-source review and assume higher slashability uncertainty.
Timing windows are stale, ambiguous, or longer than the liquidation path can safely absorb.	Reduce LTV or disallow looping until timing evidence is current.
Insurance capacity is not mapped to the specific slash surface.	Do not count it as coverage for that exposure.

Those responses are policy framing, not an implemented parameter model. Later protocol work can turn them into formulas only after every input has an observable path, freshness rule, and failure-mode treatment.

Source posture

AVS-risk claims in this section were checked against current primary EigenLayer and Symbiotic documentation. The most important source constraints are:

EigenLayer Operator Sets determine which operators secure an AVS, how allocated Unique Stake becomes slashable, and how slashed funds can be burned or redistributed.
Symbiotic networks, vaults, operators, slashers, epochs, and burners make the guarantee window and settlement path explicit; withdrawals can remain slashable through the relevant epoch boundary.
Both ecosystems make timing and process part of the risk surface. A collateral policy that ignores capture time, safety delays, veto or review windows, and settlement routing is incomplete even when the price feed is fresh.

Source notes are intentionally concise in MDX; exact URLs and access dates are recorded in the session report.