Skip to content
PROTOCOL RESEARCH

Pendle risk

The contracts holding the showcase's money are verifiably immutable and ownerless on both relevant chains, but the timelock question two prior reports left open is now settled and the answer is that there is none, under a 3-of-5 Safe that has already paused the whole protocol once.
SolidResearched Aug 5, 2026

Scorecard

strong: Audits & contractsEight audit folders on the main codebase (Spearbit, ChainSecurity, Dedaub, Ackee, Dingbats, WatchPug, 0xleastwood, CMichel), each verified present in Pendle's own repo, plus a $1M Cantina bounty; sPendle and market V6/V7 are thinner at one and two reports.
watch: Admin controlGovernance is a Safe v1.4.1 at 3-of-5 with zero modules and an empty guard, acting through a call router with no schedule or delay function anywhere in its ABI, so three signatures execute instantly.
strong: OracleIn-house PT/LP TWAP oracle, no external price feed consumed in the core path, and maturity redemption never consults it.
strong: Liquidity & exitA live full-exit quote on the owned position costs 0.009% impact and $0.25 against $3.22M of market liquidity; PT redeems 1:1 on 24 Sep.
strong: Yield (real vs emission)The 2.28% fixed leg is the discount to par funded by kHYPE's 1.98% staking yield, not by PENDLE emissions.
watch: Holder concentrationUnmeasured for a third month running (Nansen 403, Etherscan top-holders is Pro-only); 38.9% of supply sits outside circulation.
strong: Track recordNo protocol-level exploit as of 2026-08-05, checked line by line against DeFiLlama's 612-entry hacks list; the one adjacent entry is Penpie, a protocol built on Pendle.
🟢 strong🟡 watch / caveat🔴 weak / fund-loss risk
Verdict is a gate (worst flaw wins), not an average. Our read, not financial advice.

Live facts

auto-sourced now
TVL$1.17B
30d↓1%
Audits2
Last hacknone

DeFiLlama + our exploits feed. Cross-check the dated report against today.

TL;DR

Pendle splits a yield-bearing asset into PT (principal, redeems 1:1 at maturity) and YT (the yield stream). This run settles the question the last two reports left open, and the answer is the unflattering one. There is no timelock. Pendle's Ethereum governance address is a Safe v1.4.1 at 0x8119EC16F0573B7dAc7C0CB94EB504FB32456ee1 running 3 of 5, with zero modules enabled and an empty guard slot. The governance proxy those signatures act through is a role-gated call router whose ABI contains no schedule, queue or delay function of any kind, only aggregate() for arbitrary batched calls and an explicit pause(). Three signatures execute immediately, and in September 2024 that power paused every Pendle contract for about seven hours. Against that sits the thing that actually protects this position: PT-kHYPE and its HyperEVM market both read zero at the EIP-1967 implementation and admin slots and revert on owner(), so nobody can upgrade or reconfigure them. TVL is $1.193B, up 26.6% since the June report. No protocol-level exploit as of today.

Two corrections to the record. The multisig is 3 of 5, not the 2 of 4 that public write-ups and the last two versions of this report repeated. And the HyperEVM governance Safe, the one that matters for the chain where this money actually sits, was read this run for the first time rather than assumed: same five signers, same 3-of-5 threshold, no modules, no guard.

Checklist

Audits & contracts. Verified against Pendle's own audit tree rather than a marketing page. The main codebase directory carries eight audit folders: Spearbit-2024, ChainSecurity-2024, Dedaub, Ackee, Dingbats, WatchPug, 0xleastwood and CMichel. Every firm named in this report was checked to exist in that tree; none was carried over from secondary coverage. A Cantina bug bounty pays up to $1,000,000 for critical findings, capped at 10% of funds directly affected, live since 14 June 2024. Newer components are thinner: sPendle has one report (WatchPug), marketV6 and marketV7 two each (HickupHH3, WatchPug). That thinness sits on the token and staking layer, not on the PT this portfolio holds. On-chain, PT-kHYPE-24SEP2026 at 0x50fc4edc6346f36993bb30fe60e932504ed17391 and its market at 0xb48b0c95b2ddc464484305b7363fad5bd5b7a683 both return 0x0 at the EIP-1967 implementation and admin slots and revert on owner(). The PT contract self-reports as "PT Kinetiq Staked HYPE 24SEP2026". That is the immutability claim checked rather than repeated.

Admin control. Pendle's own deployment file names 0x8119EC16F0573B7dAc7C0CB94EB504FB32456ee1 as Ethereum governance. It is a Safe proxy on singleton 0x41675C099F32341bf84BFc5382aF534df5C7461a, which self-reports version 1.4.1. getThreshold() returns 3, getOwners() returns five addresses, getModulesPaginated returns an empty array, and the guard storage slot is zero. No Zodiac Delay or equivalent sits in the path. The proxy admin at 0xA28c08f165116587D4F3E708743B4dEe155c5E64 answers owner() with that Safe directly, so upgradable periphery can change with no notice period. The market factories on both Ethereum and HyperEVM answer to the governance proxy 0x2aD631F72fB16d91c4953A7f4260A97C2fE2f31e, and that proxy is where the timelock question is finally closed: it is an ERC1967 upgradeable proxy over a PGovernanceProxy implementation whose full ABI offers role management, scoped selector access, aggregate(), pause(address[]) and upgradeTo, and nothing resembling a delay. The 3-of-5 Safe holds DEFAULT_ADMIN_ROLE on it, confirmed by direct call. The separate dev multisig at 0xE6F0489ED91dc27f40f9dbe8f81fccbFC16b9cb1 runs 2 of 5 and shares exactly two signers with governance; it does not hold DEFAULT_ADMIN_ROLE on the governance proxy, but whether it holds a scoped or guardian role is refresh. The HyperEVM governance Safe 0x7877AdFaDEd756f3248a0EBfe8Ac2E2eF87b75Ac was read this run and matches Ethereum exactly: 3 of 5, v1.4.1, the same five owner addresses, zero modules, empty guard.

Oracle. Pendle prices PT and LP through its own on-chain TWAP oracle, deployed as pyYtLpOracle at 0x5542be50420E88dd7D5B4a3D488FA6ED82F6DAc2 and confirmed to carry bytecode at the identical address on both Ethereum and HyperEVM. Recommended TWAP window is 900 or 1800 seconds. No external price feed is consumed in the core path. The previous version of this report credited Chainlink; that claim is withdrawn, and the likely source of the error is worth naming, because Pendle ships a PendleChainlinkOracleFactory that wraps its own TWAP oracle behind a Chainlink-compatible interface for integrators. That is an adapter Pendle exports, not a feed Pendle depends on. The documented footgun is that market oracles are not initialized by default and an integrator must grow observation cardinality before reading them, a risk that lands on lending protocols taking PT as collateral. Redemption at maturity does not consult the oracle at all, so the showcase position never touches it.

Liquidity & exit. Measured this run, not assumed. A fresh quote sized to the real position, selling 69.275 PT-kHYPE for about $3,958, returns 0.009% price impact and $0.25 in fees against $3,217,730 of market liquidity, and the endpoint flags it safe to execute. The position is 0.12% of the pool. Maturity is 24 September 2026, fifty days out, and PT redeems 1:1 then no matter how thin the secondary market gets. The honest caveat is scale. All of Pendle on HyperEVM is $23.1M, which is 1.9% of protocol TVL, spread across five active markets. That is comfortable for four thousand dollars and would not be for forty thousand, and PT depth usually thins as expiry approaches.

Yield: real vs emission. The market prints 2.28% implied APY against 1.98% underlying kHYPE yield. The fixed leg is real: it is the discount to par, funded by Kinetiq's staking yield, not by PENDLE emissions. The market does carry a boosted rate of 2.82% against an aggregated 2.35%, so an incentive layer exists, but the 2.28% does not depend on it. Worth saying flatly that 2.28% is a thin fixed yield. This leg is in the portfolio as a hedge against a HYPE short; the strategy's return comes from funding, not from here.

Holder concentration. Not obtained. Nansen returned 403 and the Etherscan top-holders endpoint requires Pro, so distribution is refresh for the third report running. What is confirmed: 172.07M PENDLE circulating against 281.53M total, so 61.1% is in circulation and 38.9% of supply sits outside it, at a $238.1M market cap and $389.6M fully diluted. Reports of roughly a third of supply staked as sPENDLE come from secondary coverage of the January migration and were not verified on-chain today. None of this touches PT principal.

Track record. The no-protocol-level-exploit record still holds as of 5 August 2026. DeFiLlama's hacks list was pulled and searched entry by entry: 612 records, and the only one that surfaces on a Pendle query is Penpie. That entry is worth stating plainly rather than filtering out. On 3 September 2024 Penpie, built on top of Pendle, lost $27M to a reentrancy attack that used a malicious market created through Pendle's permissionless listing path, and DeFiLlama files it under the Magpie ecosystem, not under Pendle. Pendle's own contracts were not breached. Pendle's monitoring flagged the malicious contract, the team was on alert by 17:46 UTC, all contracts were paused, and normal operations resumed at 00:50 UTC, with roughly $105M reported safeguarded. That is the pause power from the admin section, demonstrated in production, and it is the strongest single argument that the missing timelock is a deliberate trade rather than an oversight.

Worst case

Three of five Safe signers, compromised or coerced, act with no delay to absorb the news. They cannot drain PT already sitting in an immutable market and they cannot block its maturity redemption. They can swap upgradable periphery so anyone routing through the standard path gets bad execution, and they can pause the protocol as they demonstrably did in 2024. Concretely for the showcase: the HYPE short on Pacifica keeps repricing while the PT-kHYPE hedge is frozen and unsellable, leaving the short effectively naked for the duration. The hedge would still redeem on 24 September. Between now and then it might not be exitable at all. There is no insurance fund.

A second surface deserves flagging, and it is standing rather than new. On-chain PT/YT market creation on Pendle is already permissionless and has been for years; the market-deploy function is not owner-gated, and that is precisely the surface Penpie's attacker used. The 23 July roadmap adds curation infrastructure around that surface for external teams rather than opening something previously closed. The exposure here is indirect: a malicious market cannot touch an existing immutable one, but it can damage protocols that integrate Pendle carelessly, and a repeat of 2024 would likely trigger another protocol-wide pause. Position sizing, the short maturity, and interacting with the market contract rather than the router are the mitigations that exist.

Bottom line

Solid, and less comfortably than last month. Nothing broke. No exploit, TVL up, and the exact contracts holding this portfolio's money verified immutable and ownerless on both Ethereum and HyperEVM rather than taken on trust. What changed is knowledge, not facts. The timelock two prior reports flagged as unverified does not exist, and this run went past the Safe to the governance proxy behind it to confirm that, so the absence is now established rather than merely unobserved. The threshold is 3 of 5, better than the 2 of 4 everyone including this report had been repeating. Those cut in opposite directions. For a $3,958 hedge held to a fifty-day maturity in contracts nobody can upgrade, no fund-loss deal-breaker. Holder concentration is now unmeasured for a third month, and that is the gap to close.

Data appendix

  • TVL: $1,192,646,311 (DeFiLlama, 2026-08-05). Up 26.6% from $942,008,703 on 2026-06-30. Down 22.5% from $1,538,546,328 ninety days ago on 2026-05-07, and at 8.9% of the $13,386,295,280 ATH set 2025-09-19. Source: api.llama.fi/protocol/pendle. Note for continuity: the June report's ~$955M corresponds to DeFiLlama's 1 July print of $953,536,621, not to its 30 June print.
  • TVL by chain: Ethereum $696.8M, Plasma $163.3M, Monad $154.8M, Arbitrum $130.3M, Hyperliquid L1 $23.1M, Binance $19.6M, Base $4.0M. Source: DeFiLlama currentChainTvls.
  • What drove the rebound: the Plasma and Monad deployments, together $318.1M, or 26.7% of protocol TVL. Pendle went live on Monad 19 June 2026 and its growth there rode weekly AUSD liquidity incentives of up to $75k. Incentive-assisted growth on young chains, not organic recovery of the Ethereum core. Source: cryptobriefing, Monad blog.
  • Audits: the repo's audits tree holds six component directories. main codebase (Spearbit-2024, ChainSecurity-2024, Dedaub, Ackee, Dingbats, WatchPug, 0xleastwood, CMichel); sPendle (WatchPug); marketV6 (HickupHH3, WatchPug); marketV7 (HickupHH3, WatchPug). LpPtoracle and LinearDiscountOracle exist but their contents were not enumerated this run. Source: github.com/pendle-finance/pendle-core-v2-public/tree/main/audits.
  • Bug bounty: up to $1,000,000 for critical, minimum $100k, capped at 10% of funds directly affected, Cantina-triaged, live since 14 June 2024. Source: cantina.xyz.
  • Admin, verified on-chain 2026-08-05 (Ethereum): governance 0x8119EC16F0573B7dAc7C0CB94EB504FB32456ee1, Safe singleton 0x41675C099F32341bf84BFc5382aF534df5C7461a, VERSION() = 1.4.1, getThreshold() = 3, getOwners() = 5 addresses, getModulesPaginated = empty array, guard slot 0x00. No timelock. proxyAdmin 0xA28c08f165116587D4F3E708743B4dEe155c5E64 owner() = that Safe. marketFactoryV6 owner() = governanceProxy 0x2aD631F72fB16d91c4953A7f4260A97C2fE2f31e. devMultisig 0xE6F0489ED91dc27f40f9dbe8f81fccbFC16b9cb1 threshold 2 of 5, two signers shared with governance, hasRole(DEFAULT_ADMIN_ROLE) on the governance proxy returns false; remaining scope refresh.
  • Governance proxy internals, verified 2026-08-05: 0x2aD631F72fB16d91c4953A7f4260A97C2fE2f31e is an ERC1967Proxy (implementation 0x53c26f7645500fea9a856bfa09e820dacf643ab9, EIP-1967 admin slot empty, UUPS). Implementation ABI exposes aggregate, aggregateWithScopedAccess, pause(address[]), modifyScopedAccess, modifySelectorAdmin, AccessControl role functions, a GUARDIAN role, and upgradeTo/upgradeToAndCall. It contains no schedule, queue, delay or eta function. hasRole(DEFAULT_ADMIN_ROLE, governanceSafe) returns true.
  • Admin, verified on-chain 2026-08-05 (HyperEVM): governance 0x7877AdFaDEd756f3248a0EBfe8Ac2E2eF87b75Ac, VERSION() = 1.4.1, getThreshold() = 3, getOwners() = the same five addresses as the Ethereum Safe, getModulesPaginated = empty array, guard slot 0x00. marketFactoryV6 0xB5CD902CbEF8461b8d6fa852f93784F090fd7BEb owner() = the same governance proxy.
  • Immutability, verified on-chain 2026-08-05 (HyperEVM): PT 0x50fc4edc6346f36993bb30fe60e932504ed17391 and market 0xb48b0c95b2ddc464484305b7363fad5bd5b7a683 both return 0x0 at the EIP-1967 implementation slot and admin slot, and revert on owner(). PT name() = "PT Kinetiq Staked HYPE 24SEP2026".
  • Oracle: Pendle in-house PT/LP TWAP oracle, deployed as pyYtLpOracle 0x5542be50420E88dd7D5B4a3D488FA6ED82F6DAc2, bytecode confirmed present at that same address on Ethereum and HyperEVM. Recommended window 900s or 1800s; market oracles not initialized by default. No Chainlink price feed in the core path; the prior report's Chainlink claim is withdrawn. Pendle does ship a PendleChainlinkOracleFactory 0x6502cda86f9110f3655512237C9FF2B9CE247c69, a Chainlink-interface wrapper over Pendle's own oracle, which is the probable origin of that error. Source: deployments/999-core.json, docs.pendle.finance oracle integration guide.
  • Owned position (this run): 69.275 PT-kHYPE-24SEP2026 on market 0xb48b0c95b2ddc464484305b7363fad5bd5b7a683, HyperEVM, underlying protocol Kinetiq. PT $57.14, underlying kHYPE $58.60, position value ~$3,958. Market liquidity $3,217,730, 24h volume $102,038, implied APY 2.28%, underlying APY 1.98%, aggregated 2.35%, max boosted 2.82%, expiry 2026-09-24 (50 days). Fresh full-exit quote: 0.009% price impact, $0.25 fee, 0.12% of pool, flagged safe to execute. Source: api-v2.pendle.finance, pendle-mcp liquidity_depth.
  • HyperEVM footprint: five active Pendle markets confirmed live (stHYPE, limUSD, vkHYPE, kHYPE, haHYPE); the chain is 1.9% of protocol TVL.
  • PENDLE token: $1.38, market cap $238.1M, FDV $389.6M, circulating 172.07M of 281.53M total (61.1%), ATH $7.50 on 2024-04-11, down 64.7% over one year. Source: CoinGecko, 2026-08-05.
  • Governance migration: landed. vePendle 0x4f30A9D41B80ecC5B94306AB4364951AE3170210 is filed under deprecated in Pendle's own deployment file, alongside feeDistributor, feeDistributorV2 and votingController. sPendle is deployed at 0x999999999991E178D52Cd95AFd4b00d066664144. sPENDLE launched 20 January 2026: 14-day withdrawal period or 5% instant-exit fee, over 80% of protocol revenue routed to buybacks for stakers, up to 4x loyalty multiplier for prior lockers decaying over two years, vePENDLE locks paused and snapshotted 29 January 2026 at 00:00 UTC. Staked-share figures are refresh. Source: deployments/1-core.json; The Block, CoinDesk and KuCoin coverage of the January launch.
  • Holder concentration: refresh (Nansen returned 403; Etherscan tokenTopHolders is a Pro endpoint).
  • Recent news (30d): H2 2026 roadmap presented 23 July by co-founder TN and growth lead Dan. Focus on RWA and institutional markets with New York issuers, tokenized ETFs and single bonds; curation infrastructure so external teams can create PT/YT markets and push PT into lending collateral; a Morpho-partnered yield treasury; and a Boros pivot toward cross-exchange funding-rate arbitrage with a peer-to-peer acceptance market. H1 2026 average daily V2 TVL was reported at roughly $1.3B. No security incidents found. Source: cryptobriefing, chaincatcher, weex.
  • Incident record: DeFiLlama hacks list pulled and searched, 612 entries, zero filed against Pendle, checked 2026-08-05. The single adjacent entry is Penpie, dated 2024-09-03, $27,000,000, classification "Protocol Logic", technique "Reentrancy", parent "magpie-ecosystem". Penpie's own post-mortem describes losses spanning Ethereum and Arbitrum. Pendle alerted at 17:46 UTC, paused all contracts, and resumed at 00:50 UTC, reporting ~$105M safeguarded. Pendle itself has not been exploited.

Maintained monthly. Methodology: DeFi Research Instruction v2.

← all protocols