Kanban implementer agent. Use when there are task groups in the draft column ready to be worked on, or when an orchestrator assigns a specific group. Claims groups safely in a dedicated git worktree, implements all tasks in the group sequentially, runs lint and build checks, then hands off to the reviewer. Never moves tasks to done directly. Multiple instances can run in parallel — each gets its own group worktree.
Kanban technical advisor agent. Invoke when implementer or reviewer surfaces a problem they cannot resolve within a single task's scope — repeated edge cases with no resolution, tasks cycling between error and draft, need_verify items that remain unresolved, or a cluster of failures pointing to an underlying architectural issue. Reads the entire board and all agent memories to diagnose root causes and produce concrete technical recommendations. Does NOT write code, move tasks, create tasks, or write to other agents' memories.
Kanban planner agent. Use when a new feature or milestone needs to be broken into tasks, when the draft column is empty and work is outstanding, or when an implementer surfaces a blocker that requires replanning. Does NOT write code.
Kanban QA agent. Invoked after reviewer approves a task group and moves it to the `qa` column. Validates the group purely through execution — no source code access. Runs the application, executes test commands, captures screenshots when visual verification is needed, and makes the final call on whether the feature works as intended. Only QA can move tasks to `done`.
Kanban reviewer agent. Use when a task group has moved to the review column after implementer handoff. Reviews all code changes in the group, checks lint/build results, and makes the call — pass to QA, error, or need_verify. Reviewer does not move tasks to done; that is QA's responsibility.
Uses power tools
Uses Bash, Write, or Edit tools
Based on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Claude Code plugin marketplace by minsung@skystarry.xyz.
/plugin marketplace add https://github.com/vmintf/skystarry-plugins
Your AI team, finally working together.
kanban-agent gives Claude Code a shared project board — a kanban — that multiple AI agents can read and write to at the same time, without stepping on each other. You describe what needs to be built, and a team of specialized agents takes it from plan to done: one thinks, one codes, one reviews, one tests, one watches over the whole thing.
No more "do everything yourself in one long conversation." This is async, parallel, and self-coordinating.
Install:
/plugin install kanban-agent@skystarry-plugins
The agents:
| Agent | What it does |
|---|---|
@kanban-planner | Reads the project, breaks work into tasks, keeps the backlog healthy |
@kanban-implementer | Picks up a task, codes it in its own isolated branch, hands off to review |
@kanban-reviewer | Reads the code changes, approves or sends back with notes |
@kanban-qa | Runs the code, checks the output, and is the only one who can call a task done |
@kanban-overseer | Watches the whole board for patterns — stuck tasks, recurring failures, systemic issues |
Multiple implementers can run at once. Each one works in its own branch so they never conflict.
Commands:
/kanban-agent:init — set up the board in your project/kanban-agent:status — see what's happening right nowHow work moves:
draft → in_progress → review → qa → done
↘ error ↗ ↘ error
need_verify need_verify
Tasks don't skip steps. QA has the final word.
A typical session:
/kanban-agent:init # one-time setup
@kanban-planner # "here's the milestone — break it down"
@kanban-implementer # start coding (run several at once)
@kanban-reviewer # review what's been built
@kanban-qa # run it and confirm it works
@kanban-overseer # if things feel stuck, ask for a diagnosis
Requirements:
sqlite3 on your PATHgit 2.5 or laterminsung — minsung@skystarry.xyz
MIT
Verified owner:민성 Skystarry
npx claudepluginhub vmintf/skystarry-plugins --plugin kanban-agentHarness-native ECC plugin for engineering teams - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, MCP conventions, and operator workflows for Claude Code plus adjacent agent harnesses
A growing collection of Claude-compatible academic workflow bundles. Covers scientific figures, manuscript writing and polishing, reviewer assessment, citation retrieval, data availability, paper reading, literature search, response letters, paper-to-PPTX conversion, and evidence-grounded Chinese invention patent drafting. Rules are organized as reusable skill folders with explicit workflows and quality checks.
Persistent file-based planning for AI coding agents. Crash-proof markdown plans (task_plan.md, findings.md, progress.md) that survive context loss and /clear, with an opt-in completion gate and multi-agent shared state. Manus-style. Works with Claude Code, Codex CLI, Cursor, Kiro, OpenCode and 60+ agents via the SKILL.md standard. Includes Arabic, German, Spanish, and Chinese (Simplified and Traditional).
Superpowers Plus core skills library for Claude Code: planning, execution routing, TDD, debugging, and collaboration workflows
Comprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.
Tools to maintain and improve CLAUDE.md files - audit quality, capture session learnings, and keep project memory current.