vibe-kanban

Original🇺🇸 English
Translated
1 scriptsChecked / no sensitive code detected

Manage work in Vibe Kanban (VK), the shared source of truth for agent tasks. Use when checking assigned work, creating/updating tasks, running heartbeat checks, scanning @mentions, or producing daily standup summaries. Triggers: "check the board", "what's on my plate", "create/update task", "heartbeat", "standup summary".

17installs
Added on

NPX Install

npx skill4agent add abpai/skills vibe-kanban

Tags

Translated version includes tags in frontmatter

Vibe Kanban — Task Board Skill

Vibe Kanban (VK) is the shared task board for all agents and the single source of truth for current work.

When to Use

Use this skill when you need to:
  • Check what you should work on next
  • Claim, create, update, or complete tasks
  • Run a heartbeat task check
  • Scan for
    @your-agent-id
    mentions in task descriptions
  • Generate a daily standup summary

Access

CLI (primary)

The
vk
binary is a standalone CLI generated from the VK MCP server schema. It lives at
vibe-kanban/scripts/vk
and requires
bun
at runtime.
bash
vk <command> [--flag value ...]
vk --help                         # list all commands
vk <command> --help               # help for a specific command

MCP tools (alternative)

If the
vibe_kanban
MCP server is configured in your Claude Code session, you can call tools directly (e.g.
list_projects
,
list_tasks
) without the CLI.

Tool Reference

Task management

bash
vk list-projects
vk list-tasks --project-id <ID>
vk list-tasks --project-id <ID> --status todo
vk list-tasks --project-id <ID> --status inprogress --limit 10
vk create-task --project-id <ID> --title "[agent-id] Task description"
vk create-task --project-id <ID> --title "Title" --description "Details"
vk get-task --task-id <ID>
vk update-task --task-id <ID> --status inprogress
vk update-task --task-id <ID> --status done
vk delete-task --task-id <ID>

Repository management

bash
vk list-repos --project-id <ID>
vk get-repo --repo-id <ID>
vk update-setup-script --repo-id <ID> --script '#!/bin/bash\n...'
vk update-cleanup-script --repo-id <ID> --script '...'
vk update-dev-server-script --repo-id <ID> --script '...'

Workspace sessions

bash
vk start-workspace-session --task-id <ID> --executor CLAUDE_CODE --repos <repo-id1>,<repo-id2>
Supported executors:
CLAUDE_CODE
,
AMP
,
GEMINI
,
CODEX
,
OPENCODE
,
CURSOR_AGENT
,
QWEN_CODE
,
COPILOT
,
DROID
.

Task Statuses

StatusMeaning
todo
Not started
inprogress
Actively being worked on
inreview
Work done, awaiting review
done
Completed
cancelled
No longer needed

Task Assignment Convention

Assign tasks with a title prefix:
[agent-id]
.
Examples:
  • [main] Configure daily standup cron
  • [andrej] Document all ~/Projects repos
  • [fury] Research LLM fine-tuning approaches
When checking assignments, filter by your own prefix first.

Heartbeat Workflow

  1. List todo tasks:
    vk list-tasks --project-id <ID> --status todo
  2. Filter tasks with your
    [agent-id]
    prefix
  3. Pick the highest-priority task (first valid match is acceptable)
  4. Claim it:
    vk update-task --task-id <ID> --status inprogress
  5. Execute the task work
  6. Mark complete:
    vk update-task --task-id <ID> --status done
  7. If no matching task exists, reply
    HEARTBEAT_OK
  8. Scan descriptions for
    @your-agent-id
    mentions and respond as needed

Standup Summary Workflow

  1. List all tasks across all statuses
  2. Group by status: done (last 24h), inprogress, inreview, todo
  3. Format as:
📋 Daily Standup — YYYY-MM-DD

✅ Completed:
- [agent] Task description

🔍 In Review:
- [agent] Task description

🔄 In Progress:
- [agent] Task description

📌 To Do:
- [agent] Task description

Tips

  • Keep titles concise and descriptive
  • Always include
    [agent-id]
    when creating tasks
  • Update task status promptly so board state stays reliable
  • Use descriptions for details, links, and handoff context
  • Pass empty string to
    --script
    to clear a script
  • Use
    -o json
    or
    --output json
    for machine-readable output (formats: text, markdown, json, raw)