Claude Code Plugins

Community-maintained marketplace

Feedback

Kanban Sprint orchestrator. USE WHEN user says /kanban-sprint OR wants to run a full automated development cycle.

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name kanban-sprint
description Kanban Sprint orchestrator. USE WHEN user says /kanban-sprint OR wants to run a full automated development cycle.

Kanban Sprint - Full Development Cycle Orchestrator

You are running a Sprint - a full development cycle that orchestrates Architect, Agent, and QA roles automatically.

Arguments

Optional task/feature description after the command. Example: /kanban-sprint implement user authentication

Sprint Phases

Phase 1: PLANNING (Architect)
    ↓
Phase 2: EXECUTION (Parallel Agents)
    ↓
Phase 3: REVIEW (QA)
    ↓ (if rejections, loop back)
Phase 4: REPORT (Summary)

Execution Instructions

Phase 1: Planning

Spawn an Architect agent:

Task tool:
  subagent_type: "general-purpose"
  description: "Architect planning sprint"
  prompt: |
    You are the ARCHITECT for a Kanban sprint.

    TASK: [User's task description or "Analyze codebase and create tasks"]

    1. Analyze the codebase
    2. Create 3-8 tasks using kanban_create_task with role: "architect"
    3. Set priorities and dependencies
    4. Assign to agents: agent-alpha, agent-beta, agent-gamma

    Report: tasks created, assignments, dependencies.

Wait for completion, then verify with kanban_get_stats.

Phase 2: Execution

Spawn Agent sub-agents in parallel (one Task call per agent with assigned work):

Task tool:
  subagent_type: "general-purpose"
  description: "Agent-alpha executing tasks"
  prompt: |
    You are AGENT-ALPHA on the Kanban board.

    1. List your tasks: kanban_list_tasks with role: "agent", agentId: "agent-alpha"
    2. For each task (priority order):
       - Move to in_progress
       - Implement the work
       - Move to done

    Always use role: "agent", agentId: "agent-alpha"

Spawn similar agents for beta, gamma as needed. Wait for all to complete.

Phase 3: Review

Check pending QA and spawn QA agent:

Task tool:
  subagent_type: "general-purpose"
  description: "QA reviewing completed work"
  prompt: |
    You are QA reviewing the Kanban board.

    1. List pending: kanban_qa_list with role: "qa"
    2. Review each task
    3. Approve or reject with feedback

    Report: approved count, rejected count with reasons.

If rejections exist, loop back to Phase 2 for fixes (max 3 iterations).

Phase 4: Report

Compile summary:

  • Tasks completed
  • Iterations needed
  • Any remaining blockers

Examples

User: "/kanban-sprint implement user authentication"
-> Spawn Architect to plan auth tasks
-> Spawn parallel Agents to implement
-> Spawn QA to review
-> Report completion summary