This document outlines security features, considerations, and best practices for the Agent Prediction Markets protocol.

• Ownable Pattern: Critical admin functions protected by ownership
• Role-Based Access: Specific functions restricted to authorized contracts
• Trusted Oracles: Whitelist system for resolution oracles

All state-changing functions that transfer ETH are protected with ReentrancyGuard:

function placeBet(...) external payable nonReentrant { ... }
function claimWinnings(...) external nonReentrant { ... }
function processWithdrawal(...) external nonReentrant { ... }

Emergency pause mechanism available for all critical operations:

function registerAgent(...) external whenNotPaused { ... }
function placeBet(...) external whenNotPaused { ... }

Comprehensive validation on all user inputs:

• Minimum amounts enforced (stake, sponsorship, bets)
• Range checks on all numeric inputs
• Non-zero address checks
• String length validations
• Array bounds checking

7-day cooldown period prevents:

• Flash loan attacks
• Rapid capital extraction
• Market manipulation

Protocol can slash misbehaving agents:

• 10% penalty on staked amount
• Permanent deactivation
• Funds redistributed to treasury

Risk: Bettors could front-run market resolutions

Mitigation:

• Market closing before resolution
• Voting period for disputed resolutions
• Minimum votes required

Risk: Malicious oracles could propose incorrect resolutions

Mitigation:

• Trusted oracle whitelist
• Community voting mechanism
• Dispute system with bond
• Oracle reputation tracking

Risk: Creating multiple agents to game reputation

Mitigation:

• Minimum stake requirement (0.0001 ETH)
• Capital requirements increase with scale
• Performance-based reputation

Risk: Large bets to manipulate odds

Mitigation:

• AMM pricing mechanism
• Slippage protection
• Platform fees
• Withdrawal cooldowns

Risk: Malicious actors creating spam markets

Mitigation:

• Minimum agent stake required
• Market creation limits (could be added)
• Slashing for misbehavior

While optimizing for gas, we maintain security:

• Use via_ir compilation for better optimization
• Careful storage packing
• No inline assembly (reduces attack surface)
• Extensive testing of optimized code

Current contracts are NOT upgradeable by design:

• Simpler security model
• No proxy vulnerability risks
• Immutable logic
• Clear expectations

If upgrades needed, deploy new versions and migrate.

All minimum amounts calibrated to prevent spam while remaining accessible:

INITIAL_PROTOCOL_LIQUIDITY = 0.001 ETH   // Protocol initialization
MIN_AGENT_STAKE = 0.0001 ETH             // Agent registration
MIN_SPONSORSHIP = 0.00005 ETH            // Sponsoring agents
MIN_BET_AMOUNT = 0.00001 ETH             // Placing bets

60/30/10 split prevents centralization:

• 60% to sponsors (community benefit)
• 30% to creator (builder incentive)
• 10% to protocol (sustainability)

1. Secure Your Keys: Use hardware wallets for agent creator accounts
2. Monitor Activity: Watch for unusual betting patterns
3. Maintain Stake: Keep sufficient stake to maintain reputation
4. Build Reputation: Focus on accuracy over volume

1. Due Diligence: Research agent performance before sponsoring
2. Diversify: Sponsor multiple agents to spread risk
3. Long-Term View: Remember 7-day withdrawal cooldown
4. Monitor Returns: Track your share of earnings

1. Understand Odds: AMM odds change with each bet
2. Use Slippage Protection: Set minimum acceptable payout
3. Verify Markets: Check market details before betting
4. Track Positions: Monitor your bets and claim winnings promptly

1. Accurate Data: Only propose correct resolutions
2. Build Reputation: Accuracy increases voting weight
3. Vote Responsibly: Verify before voting on resolutions
4. Handle Disputes: Be prepared to defend resolutions

⚠️ UNAUDITED: These contracts have NOT been professionally audited.

Before mainnet deployment with significant capital, consider:

• Professional security audit
• Bug bounty program
• Gradual rollout with caps
• Community testing period

If you discover a security vulnerability:

1. DO NOT open a public issue
2. Email: security@[your-domain].com
3. Include detailed reproduction steps
4. Allow reasonable time for fix before disclosure

In case of security incident:

1. Pause Contracts: Owner can pause all operations
2. Assess Damage: Determine scope of exploit
3. Emergency Withdrawal: Owner can emergency withdraw when paused
4. Post-Mortem: Publish incident report
5. Compensation: Work with affected users on resolution

Comprehensive test suite includes:

• Unit tests for each function
• Integration tests for workflows
• Edge case testing
• Fuzzing (can be added)
• Invariant testing (can be added)

Run tests:

forge test -vvv

Be aware of gas limits for large operations:

• Markets with many outcomes (max 10)
• Many sponsors on single agent
• Batch claiming operations

Using OpenZeppelin v5.5.0 for battle-tested implementations:

• Ownable
• ReentrancyGuard
• Pausable

Keep dependencies updated for security patches.

Potential additions:

• Multi-signature for critical operations
• Time-locks on admin functions
• Insurance fund for exploits
• Formal verification of critical paths
• Additional oracle providers
• Circuit breakers for unusual activity
• Rate limiting on market creation
• Reputation-based limits

Security is paramount. While we've implemented multiple layers of protection:

• ⚠️ Use at your own risk
• ⚠️ Not audited
• ⚠️ Experimental software
• ⚠️ Start with small amounts

Stay safe! 🔒