Claude Review
Autonomous read-only project review & inspection pipeline, all-Claude (claude-flow's review counterpart). Fable 5 (main thread) directs — picks dimensions, issues investigation orders, and judges findings directly; Agent(model "sonnet") does broad evidence collection, Agent(model "opus") deep analysis (security fixed to opus); mechanical checks (tests/build/lint/audit) run in the main thread and…
Claude Flow
Autonomous plan-to-done pipeline, all-Claude. Fable 5 (main thread) commands — plans, decomposes, issues per-task instructions, and judges diffs directly; Agent(model "sonnet") does volume implementation and Agent(model "opus") handles hard debugging and architecture-sensitive edits. Per-task mechanical verification (diff review + tests run directly) and one atomic commit per task, nonstop to fin…
Sdd
Spec-Driven Development workflow. Deep interview → requirements (EARS) → design (+ADR) → tasks → per-task execution with verification. Use when the user starts a new feature or project, asks for spec-first development, or says "sdd", "스펙", "명세", "스펙부터".