Claude Code plugin that approximates Managed Agents' define_outcome — drop a rubric in .claude/outcome.md and a Stop hook + grader subagent loop until satisfied. - View it on GitHub