Loops HouseNEAR Security Budget

NEAR HACKATHON
Links
Team
1 member- VIOwner
Vibe Code Guy
Overview
Problem: agents that hold and move money need a settlement layer whose contracts cannot be quietly changed or drained, and "is this chain secure?" is usually answered with opinion. What it is: a pre-registered, re-runnable test of one investment thesis for NEAR: that its long-term value rests on being trusted to settle agent money, measurable as security budget relative to the value secured, compared with Solana, Sui, Aptos and Base. Three claims with kill criteria were committed, hashed, timestamped with OpenTimestamps and anchored on NEAR before the results were written. Results: C1 bounty coverage holds, with a caveat (89.1% of NEAR's top-10 TVL has a live public bounty against a peer median of 41.8%; can't tell under a stricter reading, and the peer bar is zero in two of three size bands). C2 incident record fails (NEAR lost 8.38% of its average TVL per year over 24 months against a peer median of 2.37% to 2.77%). C3 control fails (one full-access key on the Sputnik DAO factory can replace the DAO that owns intents.near; that path covers 44.0% of assessed TVL; 0% sits behind an enforced upgrade delay). How NEAR fits: read-only view calls on NEAR mainnet for 54 accounts (access keys, near-plugins roles, Sputnik DAO policies, factory info), open-source contract code read at named commits, NEAR transactions for a sourced timeline of the 1 October 2026 NEAR Intents exploit, and the thesis hash stored on NEAR through SocialDB. Everything rebuilds with make all from cached data, with no keys; CI fails if committed outputs differ from a rebuild. A static dashboard shows each verdict, the chart behind it and the query or view call that produced it. Read-only throughout: no transactions, no testing of any vulnerability.