FORESHOCK

Every protocol here carries a dated score and the share of it backed by independently verified evidence. Where that share is still thin the assessment says so, is marked Low-signal rather than given a risk band, and holds every unevidenced category at the neutral midpoint instead of assuming it safe. Every score and its reasoning are free to read; the underlying dependency, incident and audit records are part of a plan. Each reading here is fixed at the date beside it. Re-scoring never stops, but a changed reading reaches plan holders on the day it changes rather than this page, which is what a plan is for. How a score is built, how to read one, and how the rubric was tested against real exploits are all in the methodology.

Coverage grows three ways: protocols Foreshock selects at its own discretion, protocols requested through the Request coverage button, and sponsored assessments delivered on an agreed timeline. No path has a guaranteed date except sponsorship.

Scores run from 0, the safest reading, to 100, the riskiest, where 50 is the neutral midpoint every unverified category is held at. Red marks a protocol scoring 50 or above, the band called Elevated: worse than neutral on the evidence gathered. Amber marks one below it, called Moderate. The boundary is fixed. It does not move as coverage grows, and it is not fitted to any backtest result.

Marinade Liquid Staking

Marinade Liquid Staking official site N/A

Solana

:

  • Evidence: this reading is based on 100% of the scored weight, verified against named sources. Categories without verified data are excluded from the reading, never counted as safe.
  • Validation: measured against 2 comparable incidents/controls in this protocol's TVL, age and category bucket.
In short

Foreshock's independent reading of Marinade Liquid Staking is Moderate risk, 48.0 out of 100, as of 7 September 2026. The largest single contributor to that reading is audit profile: 8 audits/contests on record, most recent (Neodyme, 2026-05-01) has unread findings, scored as unknown, not assumed clean. All nine scored categories are backed by verified evidence. Foreshock publishes the reading and the reasoning behind it. It is not a verdict on whether to use this protocol, and it is not investment advice.

Is Marinade Liquid Staking safe?
Foreshock does not answer that as a yes or a no. Its independent reading of Marinade Liquid Staking is Moderate risk, 48.0 out of 100, as of 7 September 2026, led by audit profile: 8 audits/contests on record, most recent (Neodyme, 2026-05-01) has unread findings, scored as unknown, not assumed clean. The score and the reasoning are published in full so the reading can be checked rather than trusted.
Has Marinade Liquid Staking been audited?
8 audits/contests on record, most recent (Neodyme, 2026-05-01) has unread findings, scored as unknown, not assumed clean. An audit describes the code at the time it was reviewed. It is not a statement that the protocol is safe today.
Who controls Marinade Liquid Staking?
Admin key: 6/13 multisig (requires 6 of 13 keys to act), upgradeable without a meaningful timelock. Top 10 holders control 65.28%, adjusted figure may still include pooled custody (untagged exchanges, infrastructure contracts) due to tagging coverage limits; exclusions require positive identification.
Has Marinade Liquid Staking been exploited before?
No known own or inherited incidents. Foreshock records an incident against a protocol whether it originated there or was inherited from something it depends on.
When was this Marinade Liquid Staking assessment last updated?
7 September 2026. Readings are recomputed as evidence changes, and every category carries the date of the evidence behind it.
Score by categories
Audit profile10.1

8 audits/contests on record, most recent (Neodyme, 2026-05-01) has unread findings, scored as unknown, not assumed clean.

Dependency risk9.0

Uses 1 oracle (The direction of Marinade's oracle relationship could NOT be settled from first-party sources, and saying so is more useful than picking a side. Its mSOL token page states the price plainly as an internal ratio, 'Price of mSOL = total_staked / tokens_minted', updated after each Solana epoch, with developers reading marinade_state.msol_price or calling the conversion helpers. But the same page carries a section headed Oracles saying 'Our mSOL price is powered by Pyth and Switchboard' and announcing a Chainlink partnership. Two other first-party facts cut against an actual consumption dependency: the pure internal formula above, and the bug bounty explicitly placing 'third-party oracle data errors' OUT of scope while forbidding testing of third-party oracles. There is also no oracle account in the published address list. The assessment is that core mSOL accounting has no external price-oracle dependency and that Pyth and Switchboard publish an mSOL feed for third parties, but that is flagged rather than asserted because Marinade's own heading says otherwise. Delegation is a Stake Auction Market: validators are scored each epoch by max_yield, ranked and allocated top-down, in a last-price auction where realised yield is set by the last validator to receive stake, capped at 15 percent of TVL per validator, and validators must post a bond as collateral before receiving stake. The one named external data provider in that path is ipwhois, used for geolocation and data-centre constraints. Note a liveness question the docs raise and do not answer: the bot that updates price and delegation is permissionless, so 'anybody can run them so that the program is always operational', but no failure mode is described for the case where nobody does.), composable with 4 other protocols.

Code characteristics9.0

Admin key: 6/13 multisig (requires 6 of 13 keys to act), upgradeable without a meaningful timelock.

Governance attack surface7.5

Top 10 holders control 65.28%, adjusted figure may still include pooled custody (untagged exchanges, infrastructure contracts) due to tagging coverage limits; exclusions require positive identification.

TVL profile5.5

TVL is 87% below this protocol's own peak; TVL change (37%) is within a stable range.

Historical incidents2.3

No known own or inherited incidents.

Team factors1.8

Doxxed team, track record: unknown.

Protocol age1.5

1862 days live, past the ~1yr floor; treated the same as any older protocol, not scored progressively safer with more age.

Bug bounty1.3

Has a bug bounty program, up to $250 000, hosted on Immunefi.

Assets held (1)

Available with a plan. See pricing.

Dependencies (2)

Available with a plan. See pricing.

Incident history (0)

Available with a plan. See pricing.

Audit history (8)

Available with a plan. See pricing.

Also covered on Solana

This assessment follows Foreshock's published methodology. Exact category weights, scoring rules, thresholds, and aggregation logic are proprietary and not shown here. How to interpret this assessment

This reading will change.

This reading is fixed at the date above it, and it stays there. We keep re-scoring Marinade Liquid Staking as the evidence moves, a new audit, a changed admin control, an incident, and this page is free to read and always will be. The updated reading is what a plan buys: you are told by email on the day it changes, instead of finding out the next time you happen to look.

Marinade Liquid Staking risk score, audits and admin control · Foreshock