NEWTON / EXPLORER
TESTNET
LIVE
TASKS 24H265 policiesCOMPLIANT100.0%0 attestedOPERATORS32 signingRESTAKED0.04 NEWT—MEDIAN RESPONSE36sp99 36.0sRE-VERIFIED 24H—0 invalidatedBLOCK11,958,004Sepolia
NEWTON PROTOCOL · POLICY CLIENT

0x41de…39e6revision 2 · current

0x41de0c896e90fe7837e11f8a6cf1a17a987239e6 · 1 of 8 policies · Sepolia
POLICY ID
0xe15dec31…f3a5
REVISION
2 of 2
POLICIES
1 of 8
DISTINCT
1
RESPONSE EXPIRY
25 blocks
CHAIN
Sepolia
TASKS · 24H
0
ALLOW RATE
—
Every policy must allowFail-closed, and no short-circuit: every policy is evaluated even after one has denied, and a single denial denies the intent. The response expires 25 blocks after its reference block — the shortest expireAfter in the set, on slot 0.

Policies — 1

one Rego + WASM pair per slot
  1. 0
    0x3088b82aefa25f086889e992013633118707a9880x3088…a988not in factory registry25 blocksbounds set expiry
    wasmnone — pure Regorego
    — rules— denyno WASMspecHash 0x4619a37c…
    whitelistnot in schema0x546d2ccb9b5f8f1535010143945f9869835c55d9
    policyParams — 58 bytes on chain0x7b2277686974656c697374223a22307835343664326363623962356638663135333530313031343339343566393836393833356335356439227d
Each module is evaluated against its own input — {"params": <this slot’s policyParams>, "wasm": <this slot’s WASM output>}. There is no shared data.wasm object and no cross-slot namespace, so slot 1 cannot read slot 0’s WASM output and a policy cannot be written to depend on its neighbours.
SET IDENTITY
policyId0xe15dec3185497574d14ab03ea22ae20c452b91a0e627b9199b3b6165b265f3a5
POLICY_SET_DOMAINkeccak256("newton.policy.set")
chainId11155111
policyClient0x41de0c896e90fe7837e11f8a6cf1a17a987239e6
revision2
policies0x3088…a988@25
Task managerreads getPolicies() at task creation, then never again
policyId = keccak256(abi.encode(POLICY_SET_DOMAIN, chainId, client, revision, policies)), where policies is the complete ordered array with each entry’s parameters and expiry. Order is part of it, so a reorder alone is a new id.
WHAT A TASK FREEZES
A task carries this exact ordered set as its own snapshot, plus one WASM input per slot. Response validation reads that snapshot and never re-reads the client, so reconfiguring mid-flight cannot change what an in-flight task means — and replay proves it by recomputing the policyId over the snapshot and comparing against the task’s own.
The bound is MAX_POLICIES = 8, enforced on chain and mirrored in the operator, the enclave and the proof guest. An oversize set is rejected, never truncated, and an empty one is rejected too — there is no vacuous allow.

Configuration history

2 writes · newest first
RevPolicy setChangePoliciesConfigured byBlockTasks
2 · current0xe15dec31…f3a5identical re-apply0x3088b82aefa25f086889e992013633118707a988—11,908,5940x8601…71000
10x09846561…2a87first set0x3088b82aefa25f086889e992013633118707a988—11,907,6870x9c3f…349d0
Every write mints a new policyId and revokes every attestation bound to the previous one, because the id commits the client, the chain, the revision counter and the exact ordered list — keccak256(POLICY_SET_DOMAIN, chainId, client, revision, policies). Revision 2 is the only one whose attestations are still spendable. One write here re-applied a byte-identical list. PoliciesUpdated fires anyway and the revision advances anyway, so that write changed nothing about the policies and still killed every outstanding authorization.

Task history — every revision

no tasksExport CSVJSON
No tasks match this scope yet.
The export carries one row per task with the frozen policy set in slot order, and it is the complete filtered history rather than this page. It has no per-policy verdict column: TaskResponse.allowed is a single boolean for the whole set, so which policy denied a task is in no record.