HexCommunity

I asked Claud if Bull Bear swap is trustworthy contract. since people want to dunk on other projects. Muh 3rd party audits are worth nothing in the age of AI as we have seen. Its a common tactic for people to check the audit box. Who is doing the audit? What firm? Name them... show the receipts.. Why is it not on the webpage? The amount of bashing this guys does on @LibertySwapFi tells me he is getting paid to do so with a cover of "Concerned citizen" muh unverified contracts trope. And now Claud. For a user doing small swaps with a tight slippage floor, it’s reasonably trustworthy. For large value, or for anyone treating it as audited and finished, it isn’t there yet. The evidence for that comes mostly from the code’s own comments. What’s genuinely good The most important property holds. Every path pulls tokens either from msg.sender or from a Permit2 signer it has verified, never from a third party. So a standing approval to this contract can’t be used to drain someone who isn’t trading. That’s the failure that wipes out aggregator users, and this design rules it out structurally. Beyond that, the defensive posture is solid: Per-trade baselines, so a trade can only spend what it brought in. A hard floor on what the user nets. A fee ceiling measured against output the caller can’t inflate. Reentrancy guards, plus slot-bound callbacks. Fail-closed defaults on almost every optional path. Most bugs that remain, including the V4 hook issue I flagged, are bounded by the user’s minBuyAmount. In practice the worst case per trade is usually “you lose your slippage tolerance,” not “you lose your tokens.” What lowers my confidence Bugs were found after money was live, and the rate hasn’t clearly slowed. The comments record a lot of them: Three live-fund-theft bugs in one night (09-14). A zero-capital theft of stranded balances, confirmed on the deployed bytecode. A permanent denial of service on the live contract (RT-30). Every stable-pool leg bricked on mainnet. A first relayed fill front-run by a searcher. A facet codehash that the real build could never match. A transient-storage version that would have reverted every trade on PulseChain. Red-team rounds run to at least 13, and I found a new exploitable issue in one read-through. When a fresh pass still finds real problems, the codebase hasn’t converged. Heavy churn in a short window. The dated comments run from 08-12 to 09-19, about five weeks. Fixes are layered on fixes. At least one “fix” (RT7-A-1) was itself wrong and had to be reverted. Known-inert checks are shipping. _assertBaselinesHeld is labeled “experimental,” and one of its intended guarantees is documented as “proven false by test, kept here only as a… deliberately-inert marker.” Being honest about it is good. Shipping it is a sign the code isn’t settled. The documentation contradicts the code. The header still claims no admin key, full immutability, and a benchmark guarantee that was deliberately removed. If the top-of-file claims are stale, every other comment needs checking before you rely on it. The attack surface is large. It has eight live venue adapters and six entry points, plus a delegatecall facet, Permit2, two signature schemes and a relayer path. It sits close enough to the 24 KB contract size limit that size pressure has driven refactors. Complexity is where these bugs keep coming from. The red-team rounds look internal. Nothing references an outside audit firm, formal verification, or a public bug bounty. Your test suite, fuzzing, or an external audit may exist outside this file; I can only judge what’s here. What would move it to trustworthy for size Freeze features. Fix the open items. Get an independent audit from a firm that knows Uniswap V4 hooks and Balancer V3. Add invariant fuzzing (Foundry or Echidna) for the “holds nothing / pulls only from the payer” properties. Run a public bug bounty. Accumulate time in production with no new findings. Until then, keep per-trade floors tight and keep the guardian pause armed. That pause is exactly the right tool for a contract at this stage, even though the header says it doesn’t exist.