Loops HouseBugBounty Swarm

NEAR HACKATHON
Links
Team
1 member- HAOwner
Harsha Jain
Overview
Bug-Bounty Swarm turns security testing into an on-chain agent economy. A coordinator maps a deliberately vulnerable demo shop with a Playwright recon agent, then auctions each testing task (access control, injection, business logic) on a NEAR market contract. Five sub-account agents (three specialists, two generalists) place sealed bids: hashed commitments first, then reveals. The contract checks every reveal, scores the bids and picks the winner. The winning agent runs its test, and an independent verifier reproduces the finding: cross-user order access, stored XSS executing in real Chromium, and a $0.01 checkout. Only then does the contract release the escrowed reward to the agent's sub-account and increment its on-chain reputation. A second contract anchors hashes of every mission report and each agent's contribution, so "Verify integrity" shows MATCH, and an edited report no longer matches the on-chain hash.
Built for reproducibility: agents are deterministic and rule-based, and bidding strategies are scripted, so every run gives the same result. The commit-reveal auctions, scoring, verification-gated payouts and attestations are real NEAR testnet transactions. Judges can check live contract state with no keys: npm run verify:market and npm run verify:near.
Stack: Rust (near-sdk) contracts, Node/TypeScript, Playwright, React. Testnet only. Tracks: AI Swarms, Agent Economy, AI Security & Defense.