#46886 [SC-Low] `destroyAgent()` functionality can easily be bricked due to Frontrunning Attack

Submitted on Jun 5th 2025 at 21:15:46 UTC by @danvinci_20 for Audit Comp | Flare | FAssets

• Report ID: #46886

• Report Type: Smart Contract

• Report severity: Low

• Target: https://github.com/flare-foundation/fassets/blob/main/contracts/assetManager/implementation/CollateralPool.sol

• Impacts:

  • Griefing (e.g. no profit motive for an attacker, but damage to the users or the protocol)

Description

Description

The agent destruction process is designed to safely decommission agents by enforcing a waiting period after which the agent can now call the destroyAgent() function to finalize the whole process. Now this function makes an external call to the CollateralPool to destroy it:

 function destroy(address payable _recipient)
        external
        onlyAssetManager
        nonReentrant
    {
@>>        require(token.totalSupply() == 0, "cannot destroy a pool with issued tokens");
        token.destroy(_recipient);
        // transfer native balance, if any (used to be done by selfdestruct)
        Transfers.transferNAT(_recipient, address(this).balance);
        // transfer untracked f-assets and wNat, if any
     .............................
    }

There exist a condition that ensures no owned funds remain in the collateral pool (totalSupply == 0). This invariant is critical for maintaining the integrity and finality of the agent lifecycle.

However, a vulnerability exists in the CollateralPool contract due to the lack of a state check in the enter() function. Specifically, enter() does not verify whether the associated agent is currently in the DESTROYING state before allowing deposits.

This opens a griefing vector: an attacker can frontrun the destroyAgent() transaction by depositing a small amount of collateral via enter(). This breaks the destruction condition (since totalSupply > 0), causing the agent’s destroy() call to fail. After the failure, the attacker can simply withdraw their funds using exit().

Since this attack does not require significant capital and can be repeated indefinitely, it effectively allows malicious actors to prevent agents from ever exiting the system.

Impact Details

This attack makes it impossible for agents to safely exit the system, since the Agents are permanently or repeatedly blocked from exiting the system.

Recommendations

Add a check in the enter() function to ensure that no new deposits can be made if the agent is in the DESTROYING state:

function enter(uint256 _fAssets, bool _enterWithFullFAssets)
    external payable
    nonReentrant
    returns (uint256, uint256)
{
    Agent.State storage agent = Agent.get(agentVault);
    require(agent.status != Agent.Status.DESTROYING, "agent is being destroyed");
    
    AssetData memory assetData = _getAssetData();
    // ... rest of the logic
}

A more efficient recommendation is to not bother about the remaining total supply when destorying agent vault since it's a timelock operation and agents have made announcement already legit users should have removed their funds safely before then, most likely the remaining share supply are dead shares.

Proof of Concept

Proof of Concept

This is the attack path:

1. Agent announces destruction via announceDestroyAgent; status changes to DESTROYING and a waiting period begins.

2. After some time the agent calls the destroyAgent() to finalize the process

3. Attacker frontruns this process by calling enter() with a minimal deposit into the collateralPOOL.

4. When the agent later attempts to finalize destruction via destroy(), the call fails due to totalSupply > 0.

5. Attacker backruns and calls exit() to withdraw their deposit.