Problem
We can now verify one directory of crash reports plus reproducers locally, but there is no repository workflow that runs representative syzbot batches through the VM-backed oracle and preserves the results.
Why this matters
Before outreach, we need repeatable evidence on real reports, not just ad hoc local runs.
Scope
- add a manual
workflow_dispatch job for batch validation against a curated syzbot corpus
- optionally add a nightly job once the guest assets are stable
- persist machine-readable oracle verdicts as build artifacts
- summarize verified / reproduced / static-only counts in the job output
Acceptance
- a maintainer can trigger a batch validation run from GitHub
- artifacts include JSON results for each crash report
- the workflow clearly distinguishes VM-reproduced matches from static-only verdicts
Problem
We can now verify one directory of crash reports plus reproducers locally, but there is no repository workflow that runs representative syzbot batches through the VM-backed oracle and preserves the results.
Why this matters
Before outreach, we need repeatable evidence on real reports, not just ad hoc local runs.
Scope
workflow_dispatchjob for batch validation against a curated syzbot corpusAcceptance