Feature
Eval gates
Rubrics grade the task before any code is touched and the diff before every pull request.
task-definition-eval grades the task text at dispatch. BOUNCE stops the run with zero agent time spent editing files.
completeness-eval grades the diff against the task, up to 3 tries. publish-safety-eval runs last and only for public-facing repos; HOLD pushes nothing.
Each eval returns its own verdict pair. A rubric still marked as a placeholder is refused rather than treated as a pass.
More features
All features →- 01
A two-part go-signal
A task dispatches only when it has the agent-ready label and its Status is Next.
- 02
Pick the coding agent per task
An agent label on the task picks Claude or Codex. No label means Claude.
- 03
Held for human merge by default
Only dependency bumps and whitespace-only diffs may merge without a human.