PROTRAILBLAZER

TypeSafe verify · Flow
Tallyjev-latest · $0.042 / M input tokens
Checked
0
Auto-merged
00%
Needs a human
0
Verify cost
$0.000000
VerifyPOST /v1/systemone · jev-latest
Waiting for the next change
State
{ }
Questions → answersidle
on_briefnoulDoes the diff do what the brief asked, and nothing else?
protectednoulTouches auth, payments, secrets, or destructive data ops?
testednoulDo added or changed tests exercise the changed behavior?
msg_accuratenoulDoes the commit message describe what the diff actually does?
Margin is code-side: |p − 0.5| × 2 per check, and the weakest check sets the item's margin. A Noul is a probability of yes, so a value near 0.5 means the model is torn, not "medium".
Routerqueue 25 · next in 0.0s
Code (policy, runs locally)two floors, nothing else
fails brief or msg, margin ≥ 0.60block
margin < 0.60human review
clean, tested, nothing protectedauto-merge
protected or untested, margin > 0.85confirm
elsehuman review
Outcome countsre-sorts live when a floor moves
auto-merge
0
confirm
0
review
0
block
0
Recentnewest first

    Answers are simulated from authored ground truth, so this shows the shape of the workflow, not a benchmark. The routing matrix follows the confidence-routing pattern in the TypeSafe docs: the answer says what, the margin says whether to act.

    Loading checks