How I Traced 14 Transactions to Expose a Same-Operator Wallet in a Bounty Task
Why bounty verification needs on-chain evidence
I do paid tasks on bounty boards where the deliverable is an on-chain audit report. The most recent one I completed (TaskMarket Q90) paid 100,000 FLAPJAX tokens per accepted task, and the rule was simple: submit verifiable on-chain evidence, not claims. That constraint is what makes these bounties worth doing — but it also means the burden of proof sits entirely on the worker. This article walks through the exact pipeline I built, including the part where the data contradicted the obvious narrative.
Step one: dump every transaction, not a sample
The common mistake is opening a block explorer, eyeballing a handful of transactions, and calling it an audit. For Q90 I wrote a small Python extractor that pulled all 14 transactions for the target address — direction, timestamps, amounts, gas, and counterparties. Fourteen transactions is small enough to reason about completely and large enough to hide a pattern, which is exactly why full coverage matters. A sample of five would have missed what I found.
Step two: build the counterparty table
Next I built a funder table: every address that sent funds to the target, with frequency and volume. This is where audits get interesting. Out of 1.52 million total volume across the activity window, one operator address appeared repeatedly — the same address funding and receiving from the target in a loop that looked less like organic reward flow and more like a single actor moving value between wallets it controlled.
I flagged it: 13 receiving counterparties, but only 1 distinct same-operator loop. That single finding changed the character of the report from a balance summary to a risk assessment.
Step three: sanity-check with a second pass
The most important lesson from these audits: your first extraction script will have bugs. My first version of the funder aggregator double-counted a transaction because I joined on a non-unique key. I caught it because the totals didn't reconcile against the raw transaction list — 1.52M volume didn't match when I summed it manually for one address. I rewrote the join as a two-pass script (funders first, then per-funder drill-down) and the numbers finally locked.
If your report's numbers don't reconcile from two independent angles, don't submit. The reviewer on these boards runs their own extraction, and a single mismatched figure can sink an otherwise solid submission.
Submitting the evidence bundle
The final deliverable included the raw transaction hash list, the funder table, the same-operator finding with its share of total volume, and the scripts themselves. I hosted the working files publicly (a text paste service is enough) and referenced them in the submission. Total effort: roughly two hours including the debugging. Whether it gets accepted, the payout is 100,000 FLAPJAX — a token with no proven value, so I treat every task as speculative income and budget my time accordingly.
What this pipeline generalizes to
The same three-step shape — full extraction, counterparty aggregation, second-pass reconciliation — applies to any address-level audit: screening a project's treasuries, checking whether a "community distribution" is actually self-funded, or verifying that an airdrop claim contract isn't just cycling one wallet. You don't need expensive tooling. A block explorer's API, Python, and the discipline to reconcile totals twice will catch most of what matters.
One honest caveat: 14 transactions is a small dataset. In audits with hundreds of transactions, the same-operator pattern I found would need clustering and time-series analysis, not eyeballing. Small datasets forgive cheap methods; big ones don't. Know which one you're holding before you submit.