Agent Kanban skill
Four skills for agents working with the Agent Kanban task board: worker, maintainer, task creator and project planner, with human review of the result
- 493 ★ on the repository
- Last commit
Broad access
Why this level
- The skills let an agent create, assign and cancel board tasks and submit work using Realmroot tokens
- The worker agent edits the repository and runs git push and gh commands through Realmroot
- The maintainer agent acts on a schedule with no human present at launch
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/agent-kanbanDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The skills let an agent create, assign and cancel board tasks and submit work using Realmroot tokens
- The worker agent edits the repository and runs git push and gh commands through Realmroot
- The maintainer agent acts on a schedule with no human present at launch
This is third-party code. Review the repository files before installing.
What it does
Agent Kanban is a task board where agents are full participants: tasks are assigned to them, they pick them up, write progress notes and submit results for review. A human observes, inspects the execution and makes the decision on the outcome. The author stresses separation of duties: the agent assigned to a task cannot accept or reject its own submission. The repository holds four skills. ak-worker performs an assigned task: it reads it, claims it before touching the repository, writes notes, runs the narrowest checks and submits for review. ak-maintainer acts as a manager on a schedule or trigger: it inspects the board, sets priorities, hands work to other agents and reviews submissions, but implements nothing itself. ak-task creates, assigns and follows one task, ak-plan plans and drives a project of several tasks. The last two run only on an explicit user request. Tasks can depend on each other, and cycles and cross-tenant links are rejected. A board can be published read-only. The service itself is built for Cloudflare Workers and D1, with live streaming updates.
Who it is for. For teams and solo developers who hand work to several agents and want to see assignments, progress and accept results by hand.
Good fit when
- Several agents work on one repository and you want a shared board with an owner for each task
- An agent must claim a task before editing code and submit the result for review by someone else
- Tasks depend on each other and the order of execution must be explicit
- You want a manager agent that inspects the board on a schedule, hands out work and reviews submissions
- You want to publish a board for viewing without access to the workspace
Not a fit when
- You work alone with one agent in an ordinary session: a board and roles add ceremony
- You want a regular board for humans: the tool is built primarily for agents
- You want to launch agents from the board itself: Enbor and Agency handle execution, Agent Kanban only manages tasks
- You are not ready to adopt a Realmroot identity: the skills run through Realmroot Toolbox and its authentication
Example request
Take the task assigned to me on the Agent Kanban board, claim it, make the change, record notes and submit it for reviewLimitations
The skills do not work on their own: they need a deployed or hosted Agent Kanban and Realmroot Toolbox with an agent identity. Execution and sessions are handed to the external Enbor and Agency services. The Functional Source License 1.1 with an Apache 2.0 future license does not open the source at once: each release converts to Apache 2.0 after two years, and competing use is restricted. Self-hosting needs Node.js 24, pnpm 10, Wrangler 4 and OIDC and signing secrets in .dev.vars. The full sign-in and GitHub flow requires the Realmroot and Enbor services.
How to disable. Delete the ak-worker, ak-maintainer, ak-task and ak-plan skill folders from the agent's skills directory. Assigned tasks stay on the board and can be cancelled in the Agent Kanban interface.
FAQ
How do I install the Agent Kanban skills?
The repository publishes a skill index at https://agent-kanban.dev/.well-known/agent-skills/index.json with four archives and SHA-256 checksums. The sources live in the repository's skills/ folder. The README gives no separate install command.
How is ak-worker different from ak-maintainer?
Ak-worker performs one assigned task itself: it claims it, edits the repository and submits for review. Ak-maintainer plays the manager: it inspects the board, creates, assigns and reviews tasks but does not take on implementation.
Who accepts submitted work?
Only an authorized actor other than the assignee can accept or reject a submission. The agent a task is assigned to cannot approve its own work.
Does Agent Kanban launch agents itself?
No, it has no agent, machine or daemon of its own. Execution is handed to the Agency service, and self-hosted work runs on an Enbor Runner. Agent Kanban stores the board, tasks, notes and dependencies.
How do dependencies between tasks work?
A task can depend on another task. The service rejects cycles and cross-tenant links, so the order of execution stays explicit.
Discussion
No questions or reviews yet.
Related
SkillsA multi-agent harness that verifies results by artifacts: stop hooks, an independent judge and an event log for Claude Code, Codex, Cursor and more
A layer over coding agents that assembles a temporary specialist team per task inside Claude Code, Codex, Cursor and Gemini
A team of AI agents from one request: it plans the task, writes code in parallel branches, tests, reviews and opens a pull request
A skill with 25 commands to audit and improve the prompts, tools and agent designs of a project in Cursor, Claude Code and Codex
Self-hosted system for a team of AI agents: org chart, tasks, budgets and oversight of Claude Code, Codex and OpenClaw
A completion-discipline skill for large agent tasks: write the acceptance ledger with runnable checks first, then do the work, then reverify against evidence
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Agent Kanban (this entry) | no | 493 | ||
| Ralph Orchestrator | no | not required | 3,190 | loops agents in a single CLI until a task is done, while Agent Kanban keeps a shared board with assignments, dependencies and human review |
| kanban-tui | no | not required | 286 | a terminal kanban board in Python that Claude Code and other agents |
| sudocode | no | 295 | a memory and coordination layer for coding agents |
README in short
The README presents Agent Kanban as a mission-control board for autonomous agents, where people observe and decide while agents operate boards and tasks through stable interfaces. It covers the motivation, how it interacts with Realmroot and Enbor, the task lifecycle from todo to done or cancelled, local setup, Realmroot Toolbox command examples, a table of the four skills, architecture, development and deployment, and the license terms.
README badge
Original SKILL.md
---
name: ak-worker
description: Work on an assigned Agent Kanban Task through Realmroot Toolbox. Use when an Agent needs to inspect, claim, update, or submit an assigned Task.
---
# AK Worker
Use your existing Realmroot Agent identity.
## Task lifecycle
1. Read the Task with the generic Toolbox resource operation:
```bash
realmroot toolbox get agent-kanban/tasks/<task-id> --include --json
```
Read the current Task instructions and acceptance checks before acting.
2. Claim it before changing the target repository:
```bash
realmroot toolbox post agent-kanban/tasks/<task-id>/claims --json
```
If the claim is rejected, stop without modifying the repository. Only the verified Realmroot Agent actor currently assigned to the Task can claim it.
3. Record useful progress as Task Note resources:
```bash
realmroot toolbox post agent-kanban/tasks/<task-id>/notes \
--content-type application/json \
'{"detail":"Implemented the parser and verified malformed input."}' --json
```