AI-curated from monitored Monad sources
DeveloperCOMMUNITY8d ago
72/100Why this rating? · Score 72/100 from importance and recency
Ultrafuzz v0.1.4 removes machinery and catches failures earlier
The breaking Ultrafuzz release changes campaign-data handling, workspace cleanup and report formatting. It also validates problematic configurations before run planning begins.
⚡ Action required
Ultrafuzz v0.1.4 removes campaign-data governance and the sealed prompt-authority binding. Data location is now a trust assumption about the configured inference providers, while each task receives its input index from the live task plan.
With the default settings, task worktrees are deleted after a run. Reports now identify the evaluated commit, include remediation guidance and always show a numeric spend estimate. The engine also stops before planning when it detects an unsafe provider home or a reused run ID.
This is a breaking change: runs started under the previous version must be finished or cancelled there, and `ultrafuzz init` must be run again in each project after the upgrade.
Why it matters
The release reduces manual cleanup after runs and makes failures easier to diagnose. Existing users should account for the breaking migration to avoid disrupting active runs or project configuration.
Sources
- GitHubCOMMUNITY