How an answer is prepared
A message sends more than your words. Every answer carries a record of what else went into it.
Where to find it
When a message pulled something in, a line appears just above the answer — Used project files and earlier chats, or whatever applied. Open it for the ordered list.
It is stored on the answer, so it is still there after a reload, on a shared conversation, and on an answer from months ago.
What it can say
Applied this chat's instructionsYou set instructions on this chat. They come last, so they win against the project brief.
Applied this project's briefThe chat is in a project that has standing instructions. Every chat in the project gets them.
Read the project's shared filesThe project has files attached and this question touched them.
Read connected sourcesThe project has connected sources and they answered. The line names each one that was read.
Recalled from this project's other chatsAnother chat in the project already covered something relevant. Those passages are quoted in.
Resolved the CVEs you mentionedYou named one or more vulnerability identifiers. The records are looked up and passed in with your message.
Asked for structured findingsFindings mode was on, so the model was also asked to return its claims in a checkable shape.
When there is no line
Most messages are plain: what you wrote goes to the model and nothing else is added. Those answers show no line.
Two things are never listed, because they apply to almost every answer: the house instruction, and a model's own operating prompt from the catalog.
What it is not
It is not the model's reasoning — that is shown separately, when a model produces it. It records what went into the prompt before the model saw it.