Notes / Findings
Source: data/finding_2026-09-25-hyperliquid-cross-ecosystem-overlap.md in the repository, rendered here without changes. Source file
2026-09-25. Roadmap item 5 (broad-review round 2): scripts/lib/cross_ecosystem_overlap.py's GROUPS-based registry (used by check_cross_ecosystem_overlap.py, check_safe_modules_guards.py and check_nested_signers.py) covered Robinhood Chain, Ethereum L1, Arbitrum, Base, Tempo, Plasma and Monad -- 7 ecosystems -- but never Hyperliquid. README.md already claims this exact capability ("no cross-protocol signer-overlap detection") as a differentiator over SolGov; this closes the gap for real, internally, rather than leaving it true on paper for 7 of 10 tracked ecosystems.
An earlier internal research pass (today's competitor-gap workflow) claimed Hyperliquid needed a "key-format normalization bridge" alongside Solana and Zcash before this could be extended -- grouped as if all three were equally out of reach. Checked live before trusting it: Hyperliquid is a standard EVM chain (HyperEVM), the same 0x/Gnosis-Safe address space every other covered ecosystem already uses. There was no format barrier -- the registry had simply never been built for it. Solana (base58 pubkeys) and Zcash (shielded/transparent, non-EVM) remain genuinely out of scope; conflating Hyperliquid with them would have left a real, closeable gap undone on a false premise.
Most Hyperliquid targets are HyperCore-native (userToMultiSigSigners, a different on-chain multisig concept entirely, not readable via safe_owners_and_threshold()). Only 4 tracked targets are rooted in an actual HyperEVM Gnosis Safe -- every address below is the live root-role holder (DEFAULT_ADMIN_ROLE or RoleRegistry.owner()), read via chains/hyperliquid/scripts/methodology_test.py's own scorer functions, not guessed from a docstring:
| Group | Root Safe | Shape |
|---|---|---|
| Kinetiq staking (kHYPE + kmHYPE) | 0x18a8…f8F1 | 4-of-8 |
| para StakingVault | 0x8D23…Ca91 | 1-of-1 (a Safe, not a bare EOA) |
| Ventuals vHYPE staking | 0x7229…fE08 | 2-of-3 |
| stHYPE liquid staking | 0x97De…51F9 | 4-of-6 |
Kinetiq's own kHYPE/kmHYPE overlap (they share the SAME Safe) was already folded into crossExposureScore by chains/hyperliquid/scorers.py::_apply_hyperevm_shared_root_exposure (2026-09-21) -- that is an INTRA-Hyperliquid check. This registry answers the separate CROSS- ecosystem question: does any of these 4 Safes' signers also appear on Robinhood Chain, Ethereum L1, Arbitrum, Base, Tempo, Plasma or Monad.
Verified twice, independently: (1) a standalone comparison of all 21 unique signer addresses across the 4 Safes against 342 known signer/EOA addresses collected live from the other 7 ecosystems' registries -- zero matches; (2) the full check_cross_ecosystem_overlap.py sweep with Hyperliquid wired in (Resolved 99 groups across 8 ecosystems (..., hyperliquid: 4)) -- no signer overlap, identical-Safe-address match, or committee containment involving any Hyperliquid group. A real, checked negative, not "never looked" -- the exact distinction this project's own discipline (and check_cross_ecosystem_overlap.py's own docstring) insists on.
check_safe_modules_guards.py: none of the 4 Safes carries a module or guard. check_nested_signers.py: all 4 resolve cleanly, no unanalyzed nested signer, no unanalyzed module/guard/singleton anywhere in the now-82-Safe registry.
No crossExposureScore moves -- there was nothing to fold (the checked result is a real negative, not an unconfirmed absence). No new scoring dimension, no new calibration. Purely a registry/tooling extension: HYPERLIQUID_GROUPS in scripts/lib/cross_ecosystem_overlap.py, wired into the 3 tools that already consumed the other 7 ecosystems' registries the same way. Full test suite (2206 tests) green, unaffected.
Solana and Zcash's genuinely incompatible key formats. Hyperliquid's own HyperCore-native root keys (the majority of its tracked targets) -- a different, non-Safe multisig concept this registry's mechanism cannot read; extending coverage to those would need a different comparison method, not attempted this pass.
Not an audit, not investment advice. All notes · Back to the dashboard.