Privacy preserving ai bounty judge

[evmpapa]

Summary

Implemented the Privacy-Preserving AI Bounty Judge homework by replacing the public plaintext submission flow with a commit-reveal bounty flow.

Participants now submit only a commitment hash during the submission phase, then reveal their answer and salt during the reveal phase. The contract verifies the reveal using:

keccak256(abi.encodePacked(answer, salt, msg.sender, bountyId))

Only valid revealed answers are eligible for AI judging.

What changed

• Added commit-reveal submission flow
• Added submission and reveal deadlines
• Added one-commitment-per-user-per-bounty rule
• Added reveal validation with answer, salt, sender, and bountyId
• Excluded unrevealed submissions from judging
• Added owner-only batched judgeAll
• Added validated finalizeWinner
• Added single payout protection
• Added mock Ritual LLM output for tests
• Updated the web app flow for commit/reveal/judge/finalize
• Updated README, lifecycle docs, architecture note, and reflection answer

Testing

Verified successfully:

• npx hardhat test — 13 passing
• npm run build — passing
• npm run lint — passing

Security notes

• No private key or local .env file was committed
• Only .env.example placeholders were committed
• Sensitive local environment files remain gitignored
• The final diff was checked for private-key-shaped values