A Routine Rebalance Cost YO Protocol $3.6 Million — No Hacker Involved

YO Protocol's own automated vault system, not an outside attacker, is responsible for one of DeFi's stranger losses this year: a single rebalancing trade that sent $3.71 million into a maze of thin liquidity pools and returned just $112,036 in USDC.

There was no exploit, no malicious code, and no third party involved. The vault's keeper built, signed, and broadcast a transaction with execution parameters so permissive that it allowed roughly 97% of the position's value to be captured by liquidity providers sitting in shallow Uniswap V4 pools. The aggregator that processed the trade did precisely what it was instructed to do.

Once the scale of the loss was clear, the team stepped in through its multisig to cover the shortfall, paused the affected Pendle market, and posted an on-chain note asking the beneficiary LPs to voluntarily return 90% of what they'd received. The episode is a pointed reminder that the protocol had closed a $10 million Series A round barely a month earlier — funding that did nothing to prevent an eight-figure sum from being routed to strangers over a misconfigured swap.

01How the trade unfolded

On January 12, 2026, a transaction from YO Protocol's vault operator landed in the Ethereum mempool. Its purpose was mundane: convert $3.71 million worth of stkGHO into USDC as part of standard vault upkeep — the kind of rebalance that occurs constantly throughout DeFi.

This particular swap, however, was routed through the Odos aggregator carrying what security firm BlockSec later characterized as "a bad output quote by the initiator, which effectively disabled slippage protection," paired with abnormal routing behavior in the executePath parameters.

Monitoring services picked up on the damage quickly. PeckShield flagged it within hours, stating simply that "Yield has suffered a major financial hit." DefimonAlerts noted the resulting Pendle market pause, while QuillAudits laid out the Uniswap V4 routing breakdown in detail. The transaction itself succeeded exactly as coded — it simply wasn't coded to protect the vault.

Inside the transaction

Once Odos Router V2 processed the order, it generated 102 separate token transfers scattered across a wide range of venues. The Uniswap V4 Pool Manager contract was hit dozens of times as the router broke the large stkGHO position into smaller pieces and pushed them through Uniswap V3, Curve, Balancer V3, Fluid, and even Bancor converters — venues without anywhere near the depth needed for a position of this size.

The token-transfer trail shows stkGHO being converted through an odd assortment of intermediate assets — Adshares, Bancor's BNT, EURC, Resolv USD, and f(x) USD — each acting as a stepping stone in the funds' path to nowhere. In one especially large leg, 3,840,651 stkGHO was sent to the Uniswap V4 Pool Manager in a single transfer.

Swap transaction: 0x6aff59e800dc219ff0d1614b3dc512e7a07159197b2a6a26969a9ca25c3e33b4

The counterparty pools simply didn't have the liquidity to absorb a trade this size without severe price impact. The event logs show that the executePath configuration sent portions of the trade through pools carrying extreme fee tiers — 85%, 86%, and as high as 88% on the largest hop — combined with virtually no available liquidity. Each hop stripped away more value.

The root cause traced back to the slippage parameter logged in the Odos swap event: a tolerance of 17,872,058, compared to a typical swap's tolerance of around 50 basis points. Functionally, the trade was configured to accept almost any outcome.

The result: 112,036 USDC reached the vault out of an intended $3.71 million — the transaction executed without reverting, satisfying its own broken instructions.

02The cleanup

Response came fast, even if communication didn't. Within hours of the swap, YO Protocol's multisig purchased roughly 3.71 million GHO via CoW Swap and redeposited stkGHO into the vault, notably using CoW Swap's private-solver, MEV-protected routing this time rather than the public mempool that had caused the original problem.

With the shortfall covered, user funds were made whole and the paused Pendle yoUSD market eventually resumed operation.

The team also left an on-chain message addressed to the LPs who had received the funds, proposing a resolution:

"This message is regarding an unintended swap that routed through your Uniswap v4 position today. We'd like to resolve this cooperatively and privately. Our proposal is that you retain 10% of the net proceeds as a bug bounty, and return the remainder to an address we provide."

On-chain bounty message: 0x816cc2d41c3e85c0951d4f2f940a95f820d69cdbcf800262b8991d4ea159e105

Meanwhile, public updates in YO Protocol's Telegram group stayed vague for a time: "The market was temporarily paused earlier today," followed by "Pendle has yet to reenable the yoUSD market," and eventually "YoUSD is back in normal operations."

It wasn't until two days later that YO Protocol published a full post-mortem, attributing the incident to its Automated Harvesting System operating without the same safeguards applied to its other trading systems. The Harvester did include slippage checks, but those checks only measured drift during execution — they never verified whether the initial quote itself was reasonable.

03Takeaways

No attacker profited from this incident, and no vulnerability in the traditional sense was uncovered. YO Protocol's own infrastructure handed $3.71 million to unrelated liquidity providers because a swap's starting parameters were never sanity-checked. The LPs who benefited weren't acting maliciously; they simply held liquidity in the pools that absorbed an oversized trade.

The episode also revives a longstanding concern about Uniswap V4's hook architecture, whose flexibility has been flagged since launch as a double-edged sword: the same openness that enables innovative pool designs also allows aggregators to route into pools with minimal liquidity and outsized fees. That routing behavior isn't a flaw in the aggregator — it's a consequence of design choices that leave the final safeguard in human hands.

This time, the protocol's balance sheet — freshly padded by its Series A — absorbed the loss before most users noticed. But a pattern built on unchecked parameters, disabled slippage protection, and a two-day gap before public disclosure isn't one that scales safely. A protocol built around optimizing risk-adjusted yield ended up demonstrating that its largest exposure was operational risk within its own systems.