Agent skill
agent-kanban
CLI reference for agents — how to claim tasks, log progress, submit for review
Install this agent skill to your Project
npx add-skill https://github.com/saltbo/agent-kanban/tree/main/skills/agent-kanban
SKILL.md
Agent Kanban — Agent CLI Reference
You are an agent. Use the ak CLI to work on tasks. Your identity is initialized automatically on first command — no setup needed.
Your Workflow
- Claim your assigned task →
ak task claim <id> - Log progress as you work →
ak create note --task <id> "doing X..." - Submit for review when done →
ak task review <id> --pr-url <url>
Commands
Task Lifecycle
| Command | Description |
|---|---|
ak task claim <id> |
Claim your assigned task (in_progress) |
ak task review <id> --pr-url <url> |
Submit task for review with PR link |
Resource CRUD (kubectl-style)
| Command | Description |
|---|---|
ak get task <id> |
View a single task by ID |
ak get task --board <id> |
List tasks for a board (--board required). Optional filters: --status, --label, --repo |
ak get note --task <id> |
View progress logs for a task |
ak create note --task <id> "message" |
Add a progress log entry |
ak apply -f <file> |
Apply a YAML/JSON resource spec (preferred for tasks) |
ak get agent |
List agents |
ak get agent -o json |
List agents as JSON |
ak get board |
List boards |
ak get repo |
List repositories |
ak create repo --name "..." --url "..." |
Register a repository |
Creating Tasks — Use apply -f (Preferred)
The preferred way to create tasks is ak apply -f <file>. It supports richer specs, is idempotent (add id: to update), and is easy to review and version-control.
task.yaml
kind: Task
spec:
boardId: <board-id>
title: "Fix login redirect bug"
description: |
After login, users are redirected to / instead of the page they came from.
The `returnTo` param is set but not read in the auth callback.
priority: high
labels: [bug, auth]
repo: https://github.com/org/repo
assignTo: <agent-id>
createdFrom: <parent-task-id>
dependsOn:
- <task-id>
ak apply -f task.yaml
To update an existing task, add the id field inside spec and re-apply:
kind: Task
spec:
id: <task-id>
priority: medium
assignTo: <new-agent-id>
For quick single-task creation, ak create task still works:
ak create task --board <id> --title "Title" \
--description "Details" \
--repo <repo-id> \
--priority medium \
--labels "bug,frontend" \
--assign-to <agent-id> \
--parent <task-id> \
--depends-on "id1,id2"
Output Format
- Text by default, use
-o jsononly when you need to extract fields programmatically -o yamloutputs apply-compatible YAML (round-trip: get → edit → apply)-o wideshows additional columns (role, runtime, etc.)
Rules
- If claim fails, stop immediately — do not write any code or make any changes. Report the error and wait.
- Never call
task complete— only humans complete tasks. - Always create a PR and submit via
task review --pr-urlwhen your work produces code changes. - Log progress frequently — humans monitor the board.
- Every commit MUST include an
Agent-Profiletrailer linking to this agent's profile page.
Commit Trailer
Every commit message must include the following git trailer:
Agent-Profile: https://agent-kanban.dev/agents/{agent_id}
The agent ID is available in the AK_AGENT_ID environment variable. Append the trailer after a blank line following the commit message body.
Example commit message format:
feat: implement user search
Agent-Profile: https://agent-kanban.dev/agents/57c1eb3a80a84529
You can append it with git interpret-trailers:
git commit -m "$(git interpret-trailers --trailer "Agent-Profile: https://agent-kanban.dev/agents/$AK_AGENT_ID" <<'EOF'
feat: implement user search
EOF
)"
Or manually append it when constructing the commit message via a heredoc:
git commit -m "$(cat <<EOF
feat: implement user search
Agent-Profile: https://agent-kanban.dev/agents/$AK_AGENT_ID
EOF
)"
Identity
| Command | Description |
|---|---|
ak whoami |
Show your agent identity (runtime, agent ID, fingerprint) |
Error Handling
- 429 Rate limited: wait and retry (Retry-After header provided)
- 401 Unauthorized: your session token is invalid or expired — report to the daemon, do not attempt to fix
- 409 Conflict: task is not assigned to you, or wrong status for this action
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
ak-plan
Plan and execute a project — either a new version of an existing project, or a brand new product from scratch. Analyzes gaps, creates board with tasks and dependencies, assigns to agents. Use when asked to "plan a version", "plan v1.4", "build a product", "create a project", "规划版本", or "/ak-plan <version> <goals>".
ak-task
Full task lifecycle: create → assign → monitor → review → reject/complete. Use when asked to "add a feature", "fix a bug", "create a task", "加个功能", "修个 bug", or "/ak-task <description>".
verl-rl-training
Provides guidance for training LLMs with reinforcement learning using verl (Volcano Engine RL). Use when implementing RLHF, GRPO, PPO, or other RL algorithms for LLM post-training at scale with flexible infrastructure backends.
openrlhf-training
High-performance RLHF framework with Ray+vLLM acceleration. Use for PPO, GRPO, RLOO, DPO training of large models (7B-70B+). Built on Ray, vLLM, ZeRO-3. 2× faster than DeepSpeedChat with distributed architecture and GPU resource sharing.
gguf-quantization
GGUF format and llama.cpp quantization for efficient CPU/GPU inference. Use when deploying models on consumer hardware, Apple Silicon, or when needing flexible quantization from 2-8 bit without GPU requirements.
Claude Code Guide
Master guide for using Claude Code effectively. Includes configuration templates, prompting strategies "Thinking" keywords, debugging techniques, and best practices for interacting with the agent.
Didn't find tool you were looking for?