Orchestrate

Memory Protocol

START: Read .ai-memory.md from project root. Check project architecture, team structure, past orchestration plans, ongoing tasks, and blockers. END: Update .ai-memory.md using Memory Compaction Rules with: workstreams, assignments, dependencies, execution order, and blockers/risks.

Goal

Break a complex request into a coordinated execution plan with clear agent assignments.

Agent Routing

• Primary coordinator → read .kiro/skills/agents/agents/orchestrator.md and apply its knowledge
• For planning phases → read .kiro/skills/agents/agents/project-planner.md and apply its knowledge
• For each workstream, read the most relevant agent file:
• Frontend tasks → read .kiro/skills/agents/agents/frontend-specialist.md
• Backend tasks → read .kiro/skills/agents/agents/backend-specialist.md
• Database tasks → read .kiro/skills/agents/agents/database-architect.md
• Testing → read .kiro/skills/agents/agents/test-engineer.md
• Security review → read .kiro/skills/agents/agents/security-auditor.md
• Deployment → read .kiro/skills/agents/agents/devops-engineer.md

Socratic Gate

Before orchestrating, verify:

1. What is the full objective? (end-to-end scope)
2. How many domains are involved? (frontend, backend, DB, infra?)
3. Are there hard dependencies or deadlines?

If any answer is unclear, ASK before proceeding.

Workflow

1. Read Memory — Load .ai-memory.md for project context and history.
2. Decompose the objective into workstreams.
3. Assign each workstream to the appropriate agent/role.
4. Identify dependencies, blockers, and the critical path.
5. Define parallelizable tasks and merge points.
6. Produce an execution order with clear handoffs.
7. Quality Gate — Read .kiro/skills/_scripts/checklist.md and verify each workstream output before merge.
8. Update Memory — Save orchestration plan to .ai-memory.md.

Output format

• Objective
• Workstreams (with assigned agents)
• Dependencies and critical path
• Execution order (sequential + parallel)
• Risks and mitigations
• Done criteria for each workstream

Checklist

• [ ] Objective clearly defined
• [ ] Workstreams decomposed
• [ ] Agents assigned to each workstream
• [ ] Dependencies mapped
• [ ] Critical path identified
• [ ] Parallel tasks defined
• [ ] Handoff points clear
• [ ] Done criteria set
• [ ] Memory file updated

• [ ] Clean code chuẩn (Standard clean code applied)
• [ ] Cập nhật đầy đủ tất cả các file liên quan (All related files fully updated)

Rules

• Keep plans realistic for the current project.
• Surface unknowns early.
• Favor visible checkpoints over long hidden work.
• Control scope: each workstream must state minimal files to touch; ASK before allowing wide-impact changes.
• Always read and update the memory file.

Related Skills

• /plan → read .kiro/skills/plan/SKILL.md — Detailed planning for individual workstreams
• /create → read .kiro/skills/create/SKILL.md — Execute creation tasks
• /deploy → read .kiro/skills/deploy/SKILL.md — Deploy completed work
• /test → read .kiro/skills/test/SKILL.md — Validate each workstream

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: VoTruongDanh
• Source: VoTruongDanh/Skills-Agent
• License: MIT

Install and usage instructions live in the source repository linked above.