add-subagent-driven-developmentlisted

# Subagent-Driven Development <!-- uses: - skill: add-backend-development - skill: add-code-review - skill: add-commit - skill: add-database-development - skill: add-final-report - skill: add-frontend-development - skill: add-tasks-checklist - skill: add-ux-design - agent: architecture-agent - agent: backend-agent - agent: database-agent - agent: discovery-agent - agent: fix-agent - agent: frontend-agent - agent: reviewer-agent - script: build-ledger.sh - script: review-package.sh - script: status.sh - script: task-brief.sh - skill: add-subagent-driven-development/references/persistent-logging-and-tasks.md --> Execute a plan by dispatching named specialist agents per task, with code review after each. **Core principle:** Named agent per task + review between tasks = high quality, fast iteration. **Second principle, equally load-bearing:** the coordinator forgets. A compaction erases the answer to "is task 4 done?", and a coordinator that lost its place re-dispatches finished work. Everything that must survive a compaction lives on disk — in the **build ledger** and in `git log` — never in the conversation. > **Provider-Agnostic:** This skill describes WHAT to dispatch (intent + prompt), not HOW. Use your platform's subagent mechanism (Task tool, sub-process, agent call, etc.). --- ## Overview **vs. Executing Plans (parallel session):** - Same session (no context switch) - Fresh subagent per task (no context pollution) - Code review after each task (catch issues early)