Multi-Agent Coding Workflow

Coding pipeline: Hana → Maestro → Rin → Kira — with automated Quality Gate

User 🧑‍💻 Posts coding request Hana 🌸 Front-facing — Relays user request Maestro ⚙️ Orchestrator — Task breakdown & routing Task → Rin | Review → Kira | Report → Hana Rin 💻 Coder — Write Code Kira 🎀 QA — Quality Gate api curl endpoint, verify response e2e browser test UI cli run CLI, test args/exit codes/output patch pytest / npm test, verify no regression PASS? CRITICAL/HIGH? 🔴 FAIL Rin fixes & re-submits 🟢 PASS Maestro Review Verifies results Hana Report Delivers results back User ✅ Receives final output ← LOOP (fix → re-review) Legend External / User Front-facing (Hana) Backend / Orchestrator Coder (Rin) QA / Quality Gate (Kira) FAIL / Loop Back PASS / Forward Quality Gate Rules 🔴 FAIL → CRITICAL / HIGH issues found 🟢 PASS → Only MEDIUM / LOW or clean Loop until 🟢 PASS, then proceed to Maestro

Agent Roles

  • • 🌸 Hana — Front-facing, receives user requests
  • • ⚙️ Maestro — Orchestrator, task breakdown & routing
  • • 💻 Rin — Coder, implements code
  • • 🎀 Kira — QA, quality gate (PASS/FAIL)

Quality Gate (Kira)

  • • 🔴 CRITICAL/HIGH → FAIL → loopback to Rin
  • • 🟢 MEDIUM/LOW/NONE → PASS → proceed to Maestro
  • • Each fix requires re-review for consistency

Routing Rules (Maestro)

  • • Implementation → Rin
  • • Review → Kira (required after implementation)
  • • Delivery → Hana (returned to user)

Multi-Agent Workflow • Hana + Maestro + Rin + Kira 🎯