[google/syzkaller] 07b1b5: pkg/aflow/flow/assessment: remove Confident output

[Dmitry Vyukov]

Branch: refs/heads/gh-readonly-queue/master/pr-7220-3313c7a196dee3b889b493ba1cf6bf56105b19ae
Home: https://github.com/google/syzkaller
Commit: 07b1b51bb958e24614eea72a0c1710b7ee2b2bbd
https://github.com/google/syzkaller/commit/07b1b51bb958e24614eea72a0c1710b7ee2b2bbd
Author: Dmitry Vyukov <dvy...@google.com>
Date: 2026-05-08 (Fri, 08 May 2026)

Changed paths:
M dashboard/app/ai.go
M dashboard/app/ai_test.go
M dashboard/app/local_ui_test.go
M pkg/aflow/ai/ai.go
M pkg/aflow/flow/assessment/kcsan.go
M pkg/aflow/flow/assessment/moderation.go

Log Message:
-----------
pkg/aflow/flow/assessment: remove Confident output

For a hundred of workflows we have now, LLM never said it's not confident.
So this approach is not working. Remove the flag.
We could add a separate agent for this, but I don't think it's worth it
for assessment flows. Perhaps for patching we could do a separate agent.

Fixes #6782


Commit: 957998906fcd85fa884c83b0f011a452a167df44
https://github.com/google/syzkaller/commit/957998906fcd85fa884c83b0f011a452a167df44
Author: Dmitry Vyukov <dvy...@google.com>
Date: 2026-05-08 (Fri, 08 May 2026)

Changed paths:
M dashboard/app/ai.go

Log Message:
-----------
dashboard/app: don't require manual confirmation for assessment jobs

The results seem to be not unreasonable, and we don't have manpower
to moderate all of them in the timely fashion.
So remove the requirement for assessment.


Commit: a6689a70fde29840bd15aed6e405e7fe95e0a6aa
https://github.com/google/syzkaller/commit/a6689a70fde29840bd15aed6e405e7fe95e0a6aa
Author: Dmitry Vyukov <dvy...@google.com>
Date: 2026-05-08 (Fri, 08 May 2026)

Changed paths:
M dashboard/app/templates/ai_job.html

Log Message:
-----------
dashboard/app: fix condition when AI job correctness is shown


Commit: 64b8ff51afa30321524d371f1ec661e2940a8327
https://github.com/google/syzkaller/commit/64b8ff51afa30321524d371f1ec661e2940a8327
Author: Dmitry Vyukov <dvy...@google.com>
Date: 2026-05-08 (Fri, 08 May 2026)

Changed paths:
M dashboard/app/ai.go
M dashboard/app/ai_test.go

Log Message:
-----------
dashboard/app: auto-create security assessment jobs

They seem to produce reasonable results,
so auto-create them automatically.


Compare: https://github.com/google/syzkaller/compare/07b1b51bb958%5E...64b8ff51afa3

To unsubscribe from these emails, change your notification settings at https://github.com/google/syzkaller/settings/notifications

[Dmitry Vyukov]

Branch: refs/heads/master

Compare: https://github.com/google/syzkaller/compare/3313c7a196de...64b8ff51afa3