Summary
The release adds daily grader-audit workflows, cross-run audit clustering, and pattern analysis to its logs and audit commands. It also tightens label permissions, MCP gateway readiness, fork pull-request safe-output handling, scanner argument handling, and dependency security gates.
Why it matters
Agentic CI becomes more auditable and reproducible while reducing unsafe automation and command-injection paths.
Related coverage
- Claude Code adds restricted execution and cross-session messaging
- Claude Code adds model-switch hooks and remote subagent visibility
- Vercel adds dashboard deployment for eve agents