The total for one quiz, out of 100: five questions worth 20 points each, graded by the
model against each question's rubric.
pass mark
The Grip Score a diff needs for the commit or push to go through. passing_score in
.grip.toml, 70 by default.
rubric
The points a complete answer covers, written by the model together with the question,
before you answer. It is used for grading and is not shown in the terminal quiz, the
questions grip ask prints, or the whiteboard in defend mode.
focus
The one-word label on a question that says what it is about: what the change does
(behaviour), why it is built this way (motivation), an edge case, a risk, or how you
would test it.
verdict
One or two sentences from the grader about the quiz as a whole, shown with the score.
board
The whiteboard: your change drawn as parts and arrows, with the untrusted inputs
marked, and five questions pinned to it. grip whiteboard draws it; see
The whiteboard.
pin
A question attached to one part or one arrow of the board. Each pin has a kind: why
this way, malicious actor, data structure, where it fails, how you know, how it works.
stage
When grip runs: pre-commit (before each commit), pre-push (before each push, the
default), or manual (grip quiz and the other commands you run yourself).
difficulty
easy, normal or hard. It changes both which questions are asked and how strictly
the answers are graded.
provider
The model that writes and grades the questions: a coding agent you already have, an
API, or a local model. See Providers.