When a protocol is exploited, the document its team publishes afterwards is usually the main public account of what happened. It is also written by the party with the most to lose from the answer. That does not make it false; it makes it a claim, and like any claim it can be checked against records the team does not control – above all, the chain itself.
What a Credible Post-Mortem Contains
- A timeline with real timestamps, ideally with block numbers or transaction hashes for the key steps. Vague phrases like “shortly after” are where uncomfortable gaps hide.
- A root cause, not just a trigger. “An attacker called function X” is a trigger. Why function X allowed it, and why that was not caught, is the root cause.
- Specific impact: which assets, how much, from which contracts, and what was recovered.
- What was tried and did not work during the response.
- Dated, specific remediation, not “we are strengthening our security”.
The Evasions to Notice
- Passive voice around decisions: “a configuration change was made” rather than who approved it.
- Blaming a dependency without saying why the team relied on it unchecked.
- “Sophisticated attacker” – occasionally true, often a substitute for explaining a simple failure.
- Leading with the recovery so the loss feels resolved before it is explained.
- The unexplained gap between detection and response.
Check the Timeline Against the Chain
- Take each transaction hash in the write-up and open it on a block explorer. Confirm the timestamp, the contract called and the amounts. Our guide to reading a block explorer covers what each field means.
- Follow the outflows. Our guide to tracing funds shows how to see where stolen assets went and whether “recovered” funds actually returned.
- Compare the stated detection time with the first exploit transaction. A long gap the document does not mention is itself information.
- Where a government has attributed the incident, read its own record. Our checks on the Ronin bridge and Bybit attributions show what such records do and do not contain.
The Upgrade Question
Many incidents trace back to who can change the code. Upgradeable contracts usually use a proxy: a fixed address that forwards calls to an implementation contract, which an admin can swap. That design is fragile in known ways – storage layouts that collide after an upgrade, implementation contracts left uninitialised, admin functions that clash with user-facing ones – and there are established mitigations for each. But every technical mitigation is subordinate to one question: who can perform the upgrade, and with what delay?
A post-mortem that blames an upgrade should say who held that power, whether a timelock applied and whether the change was reviewed. If it does not, that is the first follow-up question. The same applies to multisig signers: how many keys, held by whom, and could signers see what they approved?
Was the Bug Class Known? Could It Have Been Caught Sooner?
Two questions are rarely answered directly. Was this class of failure already documented elsewhere? If so, why did existing review not catch it – and was it inside the scope of any audit the team cites? Our guide to reading an audit report shows how to check scope. And could monitoring have detected the attack earlier? A post-mortem that addresses both is unusually candid.
No Post-Mortem at All
The strongest signal is absence. A team that does not publish anything after an incident is rarely protecting a technical secret; more often it is hoping the incident fades. Record that as the finding – no receipt – rather than filling the gap with speculation.
FAQ
Should I trust a post-mortem written by the affected team?
Treat it as a claim and check its transaction hashes and timeline on-chain.
What if the post-mortem has no transaction hashes?
Ask for them, and note their absence. A timeline you cannot check is weaker evidence.
Does “funds recovered” mean users were made whole?
Not necessarily. Check which funds returned, to which addresses, and what the team says about distribution.
Not financial advice.
Not financial advice. Human-edited; AI-assisted drafting is disclosed in our editorial guidelines. Spotted an error? See how corrections work.
