Which position should FAULTLINE look through?
Paste a supported market, vault or borrower. FAULTLINE reads it on-chain at a pinned block, maps what it depends on, and keeps it as the context for every screen.
Paste a supported market, vault or borrower. FAULTLINE reads it on-chain at a pinned block, maps what it depends on, and keeps it as the context for every screen.
{{ loadNote }}
{{ db.rec }}
{{ db.noDrivers }}
{{ xr.noGraphText }}
{{ xr.howTo }}
Each line carries a transmission weight: the share of a failure that would travel along it. The model keeps the strongest path to each dependency and multiplies the weights along it.
Select a node to see its role, its failure odds and what it would cost you if it broke.
{{ xr.node.why }}
Pick a dependency on the graph or in the list, then launch. Financial results appear once the failure reaches the position.
{{ xr.result.reason }}
{{ xr.rp.body }}
FAULTLINE did not protect real funds here. This ran on a local fork with test balances; the only transactions are on chain 31337.
The recorded Countermove run in the repository covers S1 to S4 (weETH and cbBTC markets). There is no S5 journal yet, so this replay cannot be shown without inventing data. Recording it with scripts/countermove_fork_demo.py on the PT-reUSD market would enable this slot.
This event really happened. FAULTLINE uses it to check that its risk priors are built only from data available before the cutoff. USDC closed at $0.961 on 12 March 2023; the intraday low is not in the dataset. FAULTLINE did not exist at the time and protected nothing.
data/fixtures/replay_h04.json · tests/test_hardening.pyFor deposits and unlevered positions, FAULTLINE's answer is STAY, EXIT or REVIEW. To configure a policy, use a borrow position.
No keeper is watching this position and no reserve is funded. Validating a preview does not change that.
{{ arm.fk.text }}
{{ arm.dep.text }}