Exposure taxonomy

AVS exposure is not one scalar. It is a map of which services, networks, operators, vaults, collateral paths, and timing windows can impair an LRT before liquidation or insurance can react. Lyrasing uses this taxonomy to avoid treating every ETH-denominated restaking asset as the same collateral class.

Figure 1 — One LRT routes through many operators into many slash conditions. UV-violet marks one example chain a slash event would traverse.

Taxonomy

Exposure type	Definition	Policy question
Direct AVS or network exposure	The LRT route is allocated to a service or network whose rules can slash the backing position.	Which service creates the slash surface, and what work or fault is being secured?
Operator-set exposure	A defined set of operators secures the service and can be assigned slashable work.	Which operators are active, and how much stake or voting power is behind them?
Correlated operator exposure	The same operator, operator cluster, or infrastructure provider appears across multiple LRTs, AVSs, or vault routes.	Could one operator failure affect several collateral assets at once?
Shared vault exposure	Multiple networks draw security from the same vault or reusable collateral path.	Can one network’s slash or withdrawal pressure change the guarantee available to another?
Reused collateral exposure	The same economic collateral supports several commitments through delegation, strategy, or vault composition.	Is slashability isolated, or can accounting paths create hidden coupling?
Withdrawal or epoch coupling	A position remains slashable, delayed, or claim-constrained after a withdrawal request.	Can a borrower exit faster than the risk surface settles?
Insurance coupling	A reserve or pool is shared across unrelated slash surfaces.	Does the stated coverage map to this exposure, or is it consumed by another event first?

Correlation review

The taxonomy should be applied across the whole collateral set, not only one asset at a time. A single LRT with transparent AVS exposure may still be risky inside a market if the same operators, vaults, networks, or insurance pool are already backing other admitted collateral.

Useful correlation questions:

Question	Why it matters
Do two LRTs delegate through the same operator set?	A slash, outage, or key compromise may hit both positions.
Do several services depend on the same vault or collateral route?	The market may overcount security that is already committed elsewhere.
Is the same insurance pool covering several slash surfaces?	Coverage should be allocated, not assumed to be reusable without limit.
Are withdrawal windows aligned?	Simultaneous exits can turn timing risk into liquidity risk.
Is the fault domain off-chain?	Shared node infrastructure, signing stack, or operations team can create correlation even when contracts differ.

Classification output

The taxonomy produces qualitative labels, not a fake precision score:

Label	Meaning
Isolated direct exposure	The slash surface is specific, observable, and not visibly shared with other admitted collateral.
Known correlated exposure	Overlap exists, but the operator, vault, service, and timing paths are documented well enough to cap explicitly.
Opaque exposure	One or more required surfaces are missing, stale, or discretionary.
Coupled insurance exposure	The reserve may absorb this loss, but only after checking other claims on the same capacity.
No-admission exposure	The position depends on an unobservable or unbounded slash path that cannot support collateral policy.

These labels can later feed LTV bands, borrow caps, and looping limits. They should not be collapsed into a single number until the underlying inputs have freshness rules and an implementation path.

Source notes

EigenLayer documents Operator Sets as AVS-defined operator groups and Unique Stake as slashable allocation isolated to a specific Operator Set. Symbiotic documents networks, vaults, operators, stake capture, and vault accounting as separate pieces of the shared-security path. Lyrasing borrows those distinctions for methodology language only; it does not claim current support for any particular LRT, vault, AVS, or network.