Loops HouseDryRun

NEAR HACKATHON
Links
Team
1 member- AMOwner
Aman Singh
Overview
Problem: Programs like NEAR HACKATHON take one submission that cannot be edited, and the first reader checks whether the repo runs and does what the README says. Teams find out what is missing after it is too late, and a check result in a browser tab cannot be shown to anyone later.
What we built: DryRun checks a public repository against a program's published rules (the shipped rulebook is NEAR HACKATHON): license, README commands that resolve, disclosed libraries and AI use, NEAR's role, no tracked secrets, commit window, lockfile and tests, demo-post wording. It gives a content score, an anonymised estimate board of scored entries, an idea-overlap check, an AI-readiness audit and organizer triage. No npm dependencies; 74 tests.
How NEAR fits: A testnet receipt contract, dryrun-b30f-61f4e3.testnet, stores repo URL, commit SHA, result hash, time and checker version. Receipt 1 (tx 4dBZ5SxvdfrLLP3W8ZQodWxFcpbLRaGKdS5gCJ85XEw4) is a check of near-examples/rust-status-message scoring 55. A deep review is priced with a NEAR Intents 1Click quote sent with dry: true; receipt 2 (tx 6yp9MRUiHFRWaGKDNgSmUbPenTSD6msUKLkbKMT2yMVC) stores that quote hash. NEAR AI Cloud is the optional model for deep reviews.
Results and limits: The verify page reads receipts straight from testnet RPC. The hosted API on Render is an earlier deploy without CORS, so live checks from the Pages site do not complete; the page says so and shows the recorded check (score 55) with its on-chain receipt. Run locally (npm test, npm start) for the full app. No funds move; the 1Click quote is dry. No live NEAR AI call was verified (no key). The board is an estimate, not official judging.
Try it: https://dryrun-near.pages.dev