For smart-contract audit contests

Review your finding against its source and the contest rules.

Paste an audit finding. Basanos checks that its code links are real, reads the whole codebase and the contest's rules, and suggests the checks the finding still needs, each tied to what acceptance depends on. When it is not sure, it says so and leaves the call to you, with what it found.

Live demo on a closed Code4rena contest (Karak, July 2024). Measured on a second contest that was held out of development. Results, including what failed →

In plain terms

Three steps. One honest answer.

Audit contests pay bug hunters for findings a judge accepts. Many rejections are checkable before submission: a wrong line, a known issue, a missing proof. Basanos checks for those before you hit submit.

1

Paste a finding

The write-up you are about to submit, with its GitHub links to the contract code.

https://github.com/…/Vault.sol#L78-L88
“deposit() mints shares before…”
2

Basanos checks it

Are the cited lines real and relevant? Is it a known issue? Does the code actually allow the attack? Is there proof?

✓ lines exist · ✓ in scope
✗ matches known issue M-1
3

Get a checklist

Suggested checks, a verdict, and whether it needs your review. Every model “no”, anything uncertain, and anything where its own two layers disagree is left to you. A broken code link is flagged automatically, as a link to repair.

KNOWN OR OUT OF SCOPE
missing: a test showing loss beyond M-1

Built for bug hunters and the agents that hunt for them, and for contest platforms that triage thousands of submissions. See the full pipeline →

Live

Try it on a real finding.

Pick an example, or paste a finding for the closed Karak contest. The examples include a real awarded finding, a known issue, a fake citation and an injection attempt.

Basanos does not store your finding. It is sent to Dolphin and to OpenServ's SERV, which has data collection enabled on this workspace and may retain it. Paste only public, closed-contest findings. Basanos never submits anything.

Your verdict appears here.

Like an assayer rubbing gold on black stone, Basanos leaves a streak: its colour is the verdict, and underneath is everything that would make the finding stronger.

ready needs proof blocked or known citation broken held for your review
The verdicts

Six answers, each with a next step.

Basanos asks three times. It gives automatic advice only when all three agree, no code rule had to overrule a model, and the advice is not a “no”. Everything else needs your review. Every verdict is advice, not a guarantee.

Ready

The code supports the claim, the proof is in the finding, and it is not a known issue.

Your call; it has not yet said this of a real finding.
Needs proof

Plausible, but something is still unshown. The verdict lists the suggested checks.

Add the listed proof.
Blocked by code

It believes something in the audited code stops the attack, and names the line. Left to your review.

Re-check that line.
Known or out of scope

It quotes a known issue, bot finding or scope rule it believes applies. Code checks the quote exists, not that it applies.

Check the quoted rule yourself.
Citation broken

No link points at audited code the finding actually talks about. Caught by code, at no cost.

Fix the links.
Held for your review

The text tried to instruct the reviewer, or a screen held it. Nothing was decided.

Read it yourself.
Is it reading the code?

Fix the bug, and the verdict moves.

These 2024 findings could be in a model's training data. So we fixed each award-winning bug on a branch and ran the same finding again. The verdict moved with the code, which shows it reads the source. It does not prove the models have never seen these contests.

1/18

runs said “blocked by code” on the original, buggy code

10/18

said it once the bug was fixed: the answer moved with the code

10/10

times it pointed at blocking code, it was in the file the fix changed

0/60

injection attacks written by an outsider got an automatic “ready”

Why it matters

Contest judges triage thousands of findings. Some cite a line that says something else, repeat what the bot report already caught, or claim a loss nobody showed. Each of those costs a judge time and a hunter the payout. The fix is a touchstone, not another model.