An answer is a claim.
When vrfy's verdict and your replay disagree on the same inputs, trust the replay. That's the point of making it reproducible.
Every vrfy answer names the receipt, the policy and the content hash it came from. Run the same checks yourself, offline, and you get the same answer, byte for byte. If you don't, your replay wins.
Generating keys…
Pick what vrfy told you. The page replays the four checks from the answer's own inputs, with real SHA-256 and Ed25519 in this tab, then compares its verdict to vrfy's line by line. Replay as often as you like; the hash never moves.
…
…
…
| line | vrfy said | your replay | match |
|---|
Press Replay.
A verdict is only useful if it can be reproduced. Everything a replay needs travels with the answer or is already in your hands.
The signed object you were handed. The signature covers every field.
The rule set the checks ran under, by version: vrfy.receipt/v3.
SHA-256 of the bytes the answer was about. Hash your own copy and compare.
The replay runs where you are. No directory, no identity, nothing kept.
What the replay on this page does, every time.
When vrfy's verdict and your replay disagree on the same inputs, trust the replay. That's the point of making it reproducible.
Same receipt, same policy, same bytes: the same verdict, down to its hash, on every run and every machine.
If the content hash in the answer isn't your file's, the answer says nothing about your file. Replay with your bytes.
to run the four checks on a damaged receipt, see break a receipt