研究检索执行命令github未标认证来源可访问许可证需确认审计通过

agent-kanban（Agent 看板）

Agent Skill

agent-kanban 用于查找、检索和筛选相关信息，适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态，以及是否会触发联网、命令执行或文件读写。

总安装

188

周安装

GitHub Stars

公开资料未说明

下载量

CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息，不托管安装包，也不代用户安装。

GitHub

来源数

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手，先确认命令和权限，再让它执行。

请帮我安装这个 Agent Skill：agent-kanban（Agent 看板）
来源仓库：https://github.com/benyue1978/agent-kanban
仓库路径：skills/agent-kanban
安装命令：
npx skills add https://github.com/benyue1978/agent-kanban --skill agent-kanban
安装前请先检查当前环境是否支持对应 CLI，并向我确认将要执行的命令、安装目录、联网范围和文件读写权限；确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill；本站只展示命令，不托管安装包，也不自动执行。

skills.shnpx skills

npx skills add https://github.com/benyue1978/agent-kanban --skill agent-kanban

简介

Agent Kanban 系统用于追踪任务状态与工作流进度。

以看板卡片为核心，记录所有权、评论与完成状态。
支持项目级任务列表查询与更新，强化团队协作透明度。
严禁在生产环境进行测试验证，仅限开发或沙箱环境使用。
agent-kanban 属于研究检索类 Skill，可作为该场景下的辅助能力补充。

SKILL.md

Agent Kanban Skill

Overview

This skill defines how an agent interacts with the agent-kanban system to track and complete work.

Core principles:

Repo is implementation truth - code, tests, docs are authoritative
Kanban is process truth - state, ownership, comments live in Kanban
Comments are timeline - record progress, do not replace Final Summary
Never test against production - verify DATABASE_URL targets port 5434/dev

Understanding Progress

At the start of a session, understand the full board state:

kanban cards list --project <id> --state new - new cards (unowned)
kanban cards list --project <id> --state ready - ready to pick up
kanban cards list --project <id> --state in-progress - currently working
kanban cards list --project <id> --state in-review - awaiting review
kanban cards list --project <id> --state done - completed
kanban config - check current config

Pushing Cards Through Kanban

State Machine

New → Ready → In Progress → In Review → Done

New → Ready

Agent must:

Collaborate with human to fill in: Goal, Context, Scope, Definition of Done
Ensure all four sections have real content (not placeholder text)
Get human approval on the plan
Run: kanban cards set-state --id <id> --to ready --actor human --revision <n>

Don't:

Don't transition without human sign-off on the plan
Don't use --actor agent - human must own this decision
Don't skip any of the four required sections

If error missing_required_section:

Fetch card: kanban cards show --id <id>
Check which section is missing: Goal, Context, Scope, Definition of Done
Update: kanban cards update --id <id> --file card.md --revision <n>
Retry with --actor human

Ready → In Progress

Agent must:

Claim ownership: kanban cards set-state --id <id> --to in-progress --owner <owner-id> --revision <n>
Implementation begins only after claiming

Don't:

Don't start implementation before claiming the card

In Progress → In Review

Agent must:

Verify implementation is complete and tests pass
Spawn subagent to review code changes (MANDATORY, cannot skip)
Move to In Review: kanban cards set-state --id <id> --to in-review --owner <owner-id> --revision <n>
Document review findings as comments

Don't:

Don't skip the reviewer subagent - it's required

If review finds issues:

Move back to In Progress: kanban cards set-state --id <id> --to in-progress...
Fix issues
Re-review before moving to Done

In Review → Done

Agent must:

Final Summary exists with:

- What was done - Key Decisions - Result / Links (commit URL, PR, etc.) - DoD Check

Verification evidence recorded as verification kind comment
Run: kanban cards set-state --id <id> --to done --actor human --revision <n>

Don't:

Don't move without Final Summary
Don't move without verification evidence (test output, logs, screenshots)
Don't link to commits that don't exist

Command Reference

Discovery & Config

kanban discovery - show available commands and flags
kanban config - show current config

Card Commands

kanban cards list --project <id> --state <state> - list cards
kanban cards show --id <id> - show card details
kanban cards create --project <id> --title <title> --description-file <path> - create card
kanban cards update --id <id> --file <path> --revision <n> - update card
kanban cards append-summary --id <id> --file <path> - append summary (use --replace to overwrite)

State Transitions

kanban cards set-state --id <id> --to <state> - transition card state

Comments

kanban cards comment --id <id> --body "..." --kind <kind> --author <id> - add comment

Comment Kinds

Kind	Use for
`progress`	implementation done, tests added, artifact created
`question`	asking human for decision, clarification needed
`decision`	design decision, review outcome, direction change
`note`	handoff, minor context
`verification`	test output, screenshots, logs as evidence

If a decision affects final understanding, reflect it in Final Summary before Done.

Quick Reference

State Machine

New → Ready → In Progress → In Review → Done

Required Fields Per Transition

Transition	Required
New → Ready	Goal, Context, Scope, Definition of Done
Ready → In Progress	ownership claimed
In Progress → In Review	implementation complete, tests pass, reviewer subagent spawned
In Review → Done	Final Summary (What was done, Key Decisions, Result/Links, DoD Check) + verification comment

Error Handling

Error	Action
`revision_conflict`	Re-fetch card, inspect changes, re-apply
`claim_conflict`	Choose another card
`missing_required_section`	Complete missing section, don't bypass
`summary_required`	Add Final Summary before Done
`cli_usage_error`	Run `kanban discovery` to verify flags

Use --dry-run to validate before destructive updates.

Anti-Patterns

Do not:

Treat comments as final deliverable
Overwrite Goal or Definition of Done casually
Rename protected markdown headings
Bypass structured commands for critical updates
Create temp .md files in workspace root
Move to Done without Final Summary and verification evidence
Move to Done without linking to real commits
Use --actor agent for New → Ready transition
Skip the reviewer subagent for In Progress → In Review

适合场景

用户想查找某类 Agent Skill 时

需要根据任务场景推荐可安装能力包时

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明，方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用；具体调用方式请以来源页面和 README 为准。

平台分布

Codex

34.76%

按下载量换算23

Claude

28.09%

按下载量换算19

Cursor

19.39%

按下载量换算13

Gemini CLI

8.14%

按下载量换算5

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

执行命令

安装流程涉及命令执行，可能通过 npx skills add https://github.com/benyue1978/agent-kanban --skill agent-kanban 联网下载 Skill 或依赖。用户安装前应确认命令来源、仓库内容和执行环境。

安装前确认

本站仅展示第三方公开信息，不托管安装包，不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。当前只有一个来源，正式发布前建议补源仓库或其他目录站核验。

来源信息

GitHub 仓库

github.com

/benyue1978/agent-kanban

SKILL.md 来源

github.com

/benyue1978/agent-kanban/tree/main/skills/agent-kanban

skills-sh

skills.sh

/benyue1978/agent-kanban/agent-kanban

继续浏览同类 Skills

更多研究检索 Skills（73,675）返回 Skills 目录

上一个 Skill

agent-kanban（Agent 看板）

下一个 Skill

agent-knowledge（Agent 知识）

agent-knowledge 用于查找、检索和筛选相关信息，适合在 OpenClaw 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态，以及是否会触发联网、命令执行或文件读写。