Subagent-Driven Development

Execute plan by dispatching fresh subagent per task, with two-stage review after each: spec compliance review first, then code quality review. Each subagent gets its own planning directory for structured knowledge capture.

Core principle: Fresh subagent per task + per-agent planning dir + two-stage review (spec then quality) = high quality, fast iteration

NON-NEGOTIABLE: Two-Stage Review Gate

1. Spec Compliance Review — Dispatch ./spec-reviewer-prompt.md subagent
2. Code Quality Review — Dispatch ./quality-reviewer-prompt.md subagent (only after spec review passes)

A task is NOT complete until BOTH reviews return APPROVED. No exceptions — not for "simple" tasks, config changes, or thorough self-reviews.

The Task Status Dashboard in .planning/progress.md has Spec Review and Quality Review columns. Both MUST show PASS before status can be complete.