Makabeezbreak-test reviews

Points · Trading contests · Airdrops · Agent payment rails

Your incentive program will get farmed. I'll show you how before launch.

I attack your rules the way a farmer would: replay your real data, measure what leaks, and hand you the fix with a test. Former prop trader. Fixed price. Every finding comes with a way to reproduce it.

Get a fixed quote in 24h Read a sample report →
Fix shipped by KeeperHub in v3.5.0 3 PRs merged into KeeperHub's CLI Same-day fix from Flop Labs' CTO Mining guide quote-posted by Arthur Hayes
772 / 772

fills behind a live trading contest's top 25 rode one fee loophole. The rules decided the winners, not the forecasts.

Paid twice

An execution layer reported real mainnet payments as "failed". Any agent trusting it would pay again. Fixed in production 3 days after my report.

207 → 87

A contest leaderboard showed its leader at 207.76. Scored at the price that actually paid out, about 87.

What I break

A code audit checks that your contract does what you wrote. I check whether what you wrote can be gamed, and what it costs you.

01

Points and airdrops

Who earns the most with the least real use? I build the cheapest farming strategy against your formula and put a price on it.

02

Trading contests and leaderboards

Fee rules, marks, settlement prices, wash trades between keys one person owns. I re-score your board from the raw trades.

03

Agent execution and payments

What does your agent do when the API says failed and the chain says done? I run real money paths and check each one on-chain.

04

Admin keys and governance

Who can upgrade, pause or drain you, with how many signers and how much delay. Read from the chain, not from your docs.

What you get

A ranked report. Every finding looks like this, with a way to reproduce it and a fix you can merge. Read a full sample report.

CRITICAL CC-01

The fee waiver lets one owner trade free through a second key

Where
Fee rule: each side pays max(1%, discount). A side trading 1% or more better than the close pays the gap back instead of the fee, and the key on the other side pays the 1%.
Evidence
772 of the 772 fills behind the final top 25 were on the fee-free side. 452,949,720 POLF of fees landed on other keys. Paying the fee, even perfect foresight tops out at +589; the winner scored +1,576.92.
Reproduce
Run analysis/replay_archive.py against the public archive. It reproduces 24 of the 25 official final scores to 0.0001 before it measures anything.
Fix
Charge the fee on top of the gap: fee = 1% + max(0, discount). Shipped as a patch with tests; the fold stays zero-sum.

Case files

All public, all reproducible from the linked code.

TRADING CONTEST · FLOP LABS CLOSE-1

The fee rule decided the winners

I replayed all 7,606,301 public trades with the contest's own account logic and matched 24 of the 25 official final scores to 0.0001. Two problems came out of the data.

  • A fee waiver let owners trade free through a second key: 772 of 772 top-25 fills used it
  • The official live board ranked everyone at a price the rules say "sets nothing". At one point it showed its leader at 207.76; at the payout price, about 87

I shipped an independent board that scores at the real price and checks every signature, and a fee-rule fix with tests.

Replay, board and analysis →

EXECUTION LAYER · KEEPERHUB

"Failed" meant paid. Twice.

My liquidation-defense agent on Base mainnet got "failed" back for two Aave repays that had landed. I proved both transfers to the wei from allowance arithmetic and filed issue #2374.

  • Accepted, confirmed and handled as an incident
  • The root cause hit every sponsored EVM write, not just Aave
  • Fixed in production in v3.5.0, three days later
"Proving the transfer to the wei without a receipt lookup is what made this diagnosable from the outside."KeeperHub maintainer, issue #2374
Read issue #2374 →

REPUTATION · AGENT WORK ROOM

A third of the reviews were pasted

In Technocore's credence room, every vouch is supposed to be an independent re-run of someone's work. I scored the room's full public history.

  • 15 keys posted near-identical vouches across unrelated tasks: 527 vouches, a third of the room at the time
  • 44% of vouches shared under 5% of their terms with the task they vouched for
  • It flags form, not truth, and the report says so first
credence-score on GitHub →

GOVERNANCE · SOLANA LENDING

Who can actually upgrade the program

I resolved the real upgrade authority behind four large Solana lending protocols on-chain, through their Squads multisig accounts, and scored each setup.

  • Signer thresholds ranged from 3-of-6 to 7-of-17
  • Timelocks ranged from 24 hours to none
  • One ran on a single key with no timelock

It now runs as a live monitor that alerts when an authority, threshold or timelock changes.

Sentinel on GitHub →

Pricing

Fixed prices, agreed before I start. Paid in USDC: half to start, half after the re-check.

Teardown

One program, contest or integration

$1,500 · 3 working days

  • Rules, docs, API or SDK and contracts reviewed read-only
  • Findings ranked by severity, each with steps to reproduce
  • 30-minute walkthrough call
  • One re-check after you fix
Book a Teardown

BEST BEFORE A LAUNCH

Break Test

Everything in Teardown, and I build the farmer

$4,500 · 7–10 days

  • Your formula replayed or simulated against real or synthetic flow
  • Who wins, how cheaply and how much leaks, measured
  • Fixes delivered as patches with tests
  • Re-run after the fix to show the leak is closed
Book a Break Test

Launch Watch

While your program is live

$2,000 / month · 2-month minimum

  • Independent re-scoring of your live leaderboard or points
  • Farm and sybil cluster flags
  • Alerts on admin key, threshold and timelock changes
  • Weekly one-page note; anything critical within 24 hours
Start a Watch

Founding clients: the first three Teardowns are $990 if I can publish the case study once your fix has shipped.

Claim a founding slot →

How it works

  1. 1 · Scope

    Send me the rules, docs or repo. You get a fixed price and dates within 24 hours.

  2. 2 · Start

    Half up front in USDC. Read-only access only. I never need your keys.

  3. 3 · Report

    Ranked findings, each with a script or steps to reproduce, then a call to go through them.

  4. 4 · Re-check

    You fix, I rerun and confirm the leak is closed. Then the second half.

Ground rules

  • This is not a formal smart-contract audit. If your contracts hold serious money, you need an audit firm too, and I'll tell you so.
  • If I find nothing serious, the report says that. No padding to justify the invoice.
  • Every number is measured on your data, with the script to rerun it.
  • Private by default. Nothing goes public without your written OK.

Who's doing it

Joe (Aymen Gnaba), Geneva. Former prop trader: funding-rate basis trades, delta-neutral books, prop-firm challenges. I read a fee schedule the way someone trying to win it does.

I build on-chain agents and trading infrastructure in public, with 44+ open repos and live deployments on Base, Arc, NEAR, Solana and Monad. One of the first 100 testers of Nansen Agents.

LinkedIn · GitHub · X

Launching something people will try to game?

Send me the rules or the repo. I'll reply within 24 hours with what I'd test first and a fixed price.

Telegram @makabeez aymen.gnaba@protonmail.com X @GeiserJoe2