Virtuals Protocol audit details

• Total Prize Pool: $60,000 in USDT
  • HM awards: $47,800 in USDT
    • If no valid Highs or Mediums are found, the HM pool is $0
  • QA awards: $2,000 in USDT
  • Judge awards: $5,800 in USDT
  • Validator awards: $3,900 in USDT
  • Scout awards: $500 in USDT
• Read our guidelines for more details
• Starts April 17, 2025 20:00 UTC
• Ends May 7, 2025 20:00 UTC

❗️ Important notes

1. A coded, runnable PoC is required for all High/Medium submissions to this audit.
   • An example file was provided and can be run using npx hardhat test test/poc.js
   • Exception: PoC is optional (though recommended) for wardens with signal ≥ 0.68
   • The audit repo includes a basic template to run the test suite.
   • Your submission will be marked as Insufficient if the POC is not runnable and working.
2. Since this audit includes live/deployed code, all submissions will be treated as sensitive:
   • Wardens are encouraged to use the sensitive disclosure process for High-risk submissions affecting live code, to ensure timely disclosure of such vulnerabilities to the sponsor and guarantee payout in the case where a sponsor patches a live critical during the audit.
   • Submissions will be hidden from all wardens (SR and non-SR alike) by default, to ensure that no sensitive issues are erroneously shared.
   • If the submissions include findings affecting live code, there will be no post-judging QA phase. This ensures that awards can be distributed in a timely fashion, without compromising the security of the project. (Senior members of C4 staff will review the judges’ decisions per usual.)
   • By default, submissions will not be made public until the report is published.
   • Exception: if the sponsor indicates that no submissions affect live code, then we’ll make submissions visible to all authenticated wardens, and open PJQA to SR wardens per the usual C4 process.
3. Judging phase risk adjustments:
   • High- or Medium-risk submissions downgraded to Low-risk (QA) will be ineligible for awards.
   • Upgrading a Low-risk finding from a QA report to a Medium- or High-risk finding is not supported.
   • As such, wardens are encouraged to select the appropriate risk level carefully during the submission phase.

Automated Findings / Publicly Known Issues

The 4naly3er report can be found here.

The Slither report can be found here.

Note for C4 wardens: Anything included in this Automated Findings / Publicly Known Issues section is considered a publicly known issue and is ineligible for awards.

Many protocol contracts have some form of Admin or role-related execution functions. All of these higher access functions are separate multi-sig wallets such as Fireblocks, that require at least 2 of 3 approval to run

Overview

Main Activities

VIRTUAL Genesis

1. Submit a new application at AgentFactory
   a. It will transfer VIRTUAL to AgentFactory
2. Propose at VirtualGenesisDAO (action = VirtualFactory.executeApplication )
3. Start voting at VirtualGenesisDAO
4. Execute proposal at VirtualGenesisDAO , it will do following:
   a. Clone AgentToken
   b. Clone AgentDAO
   c. Mint AgentNft
   d. Stake VIRTUAL -> $`PERSONA (depending on the symbol sent to application)
   e. Create TBA with AgentNft

Submit Contribution

1. Create proposal at AgentDAO (action = ServiceNft.mint)
2. Mint ContributionNft , it will authenticate by checking whether sender is the proposal's proposer.

Upgrading Core

1. Validator vote for contribution proposal at AgentDAO
2. Execute proposal at AgentDAO, it will mint a ServiceNft, and trigger following actions:
   a. Update maturity score
   b. Update VIRTUAL core service id.

Distribute Reward

1. On daily basis, protocol backend will conclude daily profits into a single amount.
2. Protocol backend calls AgentReward.distributeRewards , triggering following:
   a. Transfer VIRTUAL into AgentReward
   b. Account & update claimable amounts for: Protocol, Stakers, Validators, Dataset Contributors, Model Contributors

Claim Reward

1. Protocol calls AgentReward.withdrawProtocolRewards
2. Stakers, Validators, Dataset Contributors, Model Contributors calls AgentReward.claimAllRewards

Staking VIRTUAL

1. Call AgentToken.stake , pass in the validator that you would like to delegate your voting power to. It will take in sVIRTUAL and mint $`PERSONA to you.
2. Call AgentToken.withdraw to withdraw , will burn your $`PERSONA and return sVIRTUAL to you.

Links

• Previous audits:
  • https://whitepaper.virtuals.io/info-hub/important-links-and-resources/security-audits
• Documentation: https://whitepaper.virtuals.io/
• Website: https://app.virtuals.io/
• X/Twitter: https://x.com/virtuals_io
• Discord: https://discord.gg/virtualsio

---

Scope

See scope.txt

Files in scope

Files out of scope

See out_of_scope.txt

Scoping Q & A

General questions

External integrations (e.g., Uniswap) behavior in scope

EIP compliance checklist

N/A

Additional context

Main invariants

N/A

Attack ideas (where to focus for bugs)

Bonding pool

Is there a way for users to pull liquidity from the bonding pool

Any risk of loss of funds

Access control related

Are all privileged actions guarded by access controls

All trusted roles in the protocol

Describe any novel or unique curve logic or mathematical models implemented in the contracts

Simple bonding curve with a y = kx

Running tests

git clone --recurse https://github.com/code-423n4/2025-04-virtuals-protocol.git
cd 2025-04-virtuals-protocol
yarn
npx hardhat test

To run code coverage

npx hardhat coverage

Miscellaneous

Employees of Virtuals and employees' family members are ineligible to participate in this audit.

Code4rena's rules cannot be overridden by the contents of this README. In case of doubt, please check with C4 staff.