Convexa has undergone formal security analysis using industry-standard tools. This document summarizes findings and mitigations.

Date: May 3, 2026
Scope: DebateEscrow.sol, ConvictionTracker.sol
Tools: Slither v0.11.5, TruffleHog, pip-audit, npm audit
Status: All critical and high-severity issues FIXED

Issue: DebateEscrow.settleSide() transferred ETH before updating state, enabling reentrancy attacks.

Mitigation: Implemented Checks-Effects-Interactions (CEI) pattern:

1. Check all conditions upfront
2. Update all state variables BEFORE external calls
3. Make external calls last

Code:

// EFFECTS: Update state first
debateActive = false;
userStakes[staker][Side.BULL] = 0;
userStakes[staker][Side.BEAR] = 0;

// INTERACT: External calls last
(bool sent, ) = payable(staker).call{value: payout}("");

Status: ✓ Fixed and verified

Issue: Constructor and setter functions didn't validate address parameters.

Mitigation: Added explicit zero-address checks:

require(_settlementExecutor != address(0), "Invalid executor");
require(newJudgeAgent != address(0), "Invalid judge");

Status: ✓ Fixed

Issue: Judge could set conviction scores outside valid range (-100 to +100).

Mitigation: Added range validation:

require(conviction >= -100 && conviction <= 100, "Invalid conviction");

Status: ✓ Fixed

Both contracts use role-based access control:

All critical functions require explicit permissions. Unauthorized calls revert.

• Staked ETH locked in contract until settlement
• Settlement transfers happen in CEI pattern
• Re-entrancy guards via state updates before calls

• Only ETH accepted for staking (no ERC20 complexity)
• Uniswap integration uses production-grade token routing

Both contracts verified onchain:

• DebateEscrow: Blockscout
• ConvictionTracker: Blockscout

Source code matches deployed bytecode.

Run audit:

pip-audit

All dependencies pinned to specific versions. No known vulnerabilities in locked set.

Run audit:

cd contracts && npm audit

No high-severity vulnerabilities. Some medium-severity advisories for optional dev dependencies.

• DEPLOYER_PRIVATE_KEY: Used only for contract deployment (once)
• AGENT_WALLET_PRIVATE_KEY: Used only for settlement transactions (KeeperHub executor)
• Never hardcoded; loaded from .env

• All keys loaded from .env at runtime
• .env excluded from version control (.gitignore)
• Testnet-only keys (no mainnet credentials in repo)

Scan repository for exposed secrets:

trufflehog filesystem . --debug

Result: No credentials found in committed files.

Test critical security functions:

cd contracts && npx hardhat test

Test end-to-end security:

pytest tests/security/

Tests cover:

• Reentrancy prevention
• Access control enforcement
• Input validation
• Token handling

1. Testnet Only: Contracts deployed on Unichain Sepolia testnet. Not audited for mainnet use.

2. Agent Trust: System assumes judge agent (Groq LLM) produces honest verdicts. Judge agent is controlled by system operator.

3. KeeperHub Trust: System assumes KeeperHub executor is trusted to broadcast transactions fairly.

4. Oracle Risk: Market data from Uniswap API trusted as-is (no price oracle). Large price movements could cause unexpected slippage.

✓ Reentrancy protection (CEI pattern)
✓ Zero-address validation
✓ Input validation (conviction range)
✓ Access control (role-based)
✓ Token handling (only ETH, no ERC20)
✓ Private key not hardcoded
✓ No exposed secrets in repo
✓ Dependencies audited
✓ Contracts verified onchain
✓ Tests pass (unit + integration)

If you discover a security issue:

1. Do not open a public issue on GitHub
2. Email security concerns to the maintainers
3. Provide reproduction steps if possible
4. Allow time for remediation before public disclosure

• Checks-Effects-Interactions Pattern
• Slither Documentation
• OpenZeppelin Security