subagent-driven-development
Use when executing implementation plans with independent tasks in the current session or facing 3+ independent issues that can be investigated without shared state or dependencies - dispatches fresh subagent for each task with code review between tasks, enabling fast iteration with quality gates
proposal-execution
Use when executing approved proposals. Researchers can execute ANY approved proposal (not just their own) to perform actions like memory cleanup, marking proposals ready for ticket/skill creation.
project-metadata-builder
Build and update project metadata registry in ${CLAUDE_PROJECTS_YML_PATH}. Use when starting a session, after project analysis, when user says "update project metadata", "refresh projects", "register project", or when project-analysis agent completes analysis.
