Algorand contract detectors

Detects contract state mutations after InnerTxnBuilder.Execute() calls. In AVM, inner transactions execute atomically, but if global state is updated after an inner call that transfers control, the called contract may re-enter and exploit stale state. This mirrors the classic reentrancy pattern adapted for Algorand's inner transaction model.

Flags OnComplete.UpdateApplication handlers that lack sender authorization. Without a check like Assert(Txn.sender() == Global.creator_address()), any account can replace the approval program TEAL bytecode, fully compromising the contract. This is a real attack vector seen in unaudited Algorand DeFi contracts.

Detects OnComplete.DeleteApplication handlers without proper auth. If any user can delete the app, all escrowed ALGO and ASAs in the app account are permanently lost. This must always be gated by creator or admin check.

Identifies payment or asset transfer transactions where Txn.close_remainder_to() or Txn.asset_close_to() is not validated against Global.zero_address(). An attacker can drain the full account balance by setting close-to to their own address. This is one of the most common Algorand exploits.

Flags transactions that do not assert Txn.rekey_to() == Global.zero_address(). Without this, a malicious transaction can rekey the contract's auth address, granting permanent control to an attacker. Every Algorand stateful contract must validate this.

Warns when Txn.group_index() is not validated, allowing atomic group reordering attacks. An attacker can submit a valid group with transactions in a different order, bypassing sequential validation logic.

AVM uint64 arithmetic wraps on overflow and panics on underflow. Detects operations like subtraction without prior bounds checking (App.globalGet(key) - amount) where amount could exceed the stored value, causing a panic or exploitable wrap-around in older TEAL versions.

Detects contracts that don't validate Txn.fee() against a maximum threshold. Without Assert(Txn.fee() <= Int(max_fee)), an attacker can set an excessively high fee to drain the contract's ALGO balance through miner fee extraction.

Warns when contract logic paths may exceed the 700 opcode budget per transaction. Complex loops, recursive Subroutine calls, or multiple byte operations can exceed the budget, causing transaction failure. Suggests using OpUp inner transactions for budget increases.

Flags business logic that depends on Global.latest_timestamp() for critical decisions. Block proposers on Algorand can manipulate the timestamp within a ±25 second window, making time-dependent logic like auction deadlines or vesting schedules exploitable.

Detects when multiple code paths write to the same global state key with different types or semantics, creating data corruption risks. Also flags contracts approaching the 64 global state key limit.

Identifies scratch space usage (ScratchVar) where values are read before being written in a given execution path, which returns a default zero value. This can lead to logic errors when the zero is used in arithmetic or conditionals.

Detects direct access to Txn.application_args[N] without asserting the argument count first. Missing bounds checks can panic execution and create denial-of-service surfaces under malformed calls.

Flags inner payment or asset transfer builders where receiver/asset_receiver is not constrained to a trusted address or argument policy. This can route funds to attacker-controlled accounts.

Warns when box read/write operations are performed without sender authorization for sensitive namespaces. Attackers may overwrite business-critical box state when access checks are absent.

Identifies ABI dispatch handlers that do not validate the first 4-byte selector before executing a branch. This increases ambiguity and may route malformed calls into unintended logic paths.