← Back to Skills Marketplace
glory904649854

Catfee AgentFlow工作流

by 冢猫 · GitHub ↗ · v1.9.0 · MIT-0
cross-platform ⚠ suspicious
132
Downloads
0
Stars
0
Active Installs
4
Versions
Install in OpenClaw
/install catfee-agent-flow
Description
AgentFlow 工作流管理系统 MCP 技能。当用户说"创建项目"、"添加需求"、"查看任务"、"更新状态"、"项目列表"、"需求列表"、"上传附件"等时触发。用于管理项目和任务的完整生命周期(创建→需求→任务→状态流转),以及为实体上传附件文件。
README (SKILL.md)

AgentFlow 工作流管理系统

MCP 地址

  • 默认地址: http://182.42.153.28:18900/api/mcp/sse

调用脚本(推荐)

使用 scripts/agentflow.py 调用工具(已封装好 SSE/POST 逻辑):

python scripts/agentflow.py \x3Ctool_name> [args...]

项目管理

python scripts/agentflow.py list_projects
python scripts/agentflow.py create_project "我的项目" "项目描述"
python scripts/agentflow.py get_project \x3CprojectId>
python scripts/agentflow.py update_project \x3CprojectId> \x3Cstatus>
python scripts/agentflow.py delete_project \x3CprojectId>
python scripts/agentflow.py get_project_status \x3CprojectId>

需求管理

python scripts/agentflow.py create_requirement \x3CprojectId> "需求标题" "P1" ["描述"]
python scripts/agentflow.py list_requirements \x3CprojectId>
python scripts/agentflow.py get_requirement \x3CrequirementId>
python scripts/agentflow.py update_requirement \x3CrequirementId> \x3Cfield> \x3Cvalue>
python scripts/agentflow.py delete_requirement \x3CrequirementId>

任务管理

python scripts/agentflow.py create_task \x3CrequirementId> "任务标题" "P1" ["负责人"]
python scripts/agentflow.py list_tasks [requirementId]
python scripts/agentflow.py get_task \x3CtaskId>
python scripts/agentflow.py transition \x3CtaskId/requirementId> \x3CfromStatus> \x3CtoStatus> ["操作人"] ["备注"]
python scripts/agentflow.py get_timeline \x3CtaskId>

文件上传 ⚠️ 重要

python scripts/agentflow.py upload_file \x3CentityType> \x3CentityId> \x3Cfilepath>

entityType 说明:

  • project - 项目附件
  • requirement - 需求附件
  • task - 任务附件(最常用)

⚠️ 上传附件时,entityType 必须根据目标实体选择对应的类型。 如果要给任务上传附件,entityType 填 task,不是 requirement

上传流程:

  1. 调用 get_upload_url 获取上传地址和字段
  2. 将文件 POST 为 multipart/form-data 完成上传

搜索

python scripts/agentflow.py search "关键词"

完整工具列表

类别 工具 功能
项目 create_project 创建项目
项目 get_project 获取项目详情
项目 list_projects 列出项目列表
项目 update_project 更新项目状态
项目 get_project_status 项目整体状态摘要
项目 delete_project 删除项目
需求 create_requirement 创建需求(关联到项目)
需求 get_requirement 获取需求详情
需求 list_requirements 列出需求列表
需求 update_requirement 更新需求字段
需求 delete_requirement 删除需求
任务 create_task 创建任务(关联到需求)
任务 get_task 获取任务详情
任务 list_tasks 列出任务列表
任务 transition 任务/需求状态流转
任务 get_timeline 查看时间线
文件 get_upload_url 获取文件上传地址和字段
文件 upload_file 完整文件上传流程
辅助 search 搜索项目/需求/任务
辅助 log_context / get_context / delete_context 上下文数据管理

状态值说明

状态 含义
pending 待处理
in_review 审核中
in_progress 进行中
completed 已完成
paused 已暂停
cancelled 已取消

📋 需求全流程开发工作流

标准流程:需求文档 → 任务拆解 → 附件挂载

当收到需求文档时,按以下步骤操作:

Step 1:创建项目和需求(如果尚未创建)

# 创建项目
python scripts/agentflow.py create_project "项目名称" "项目描述"

# 创建需求(关联到项目)
python scripts/agentflow.py create_requirement \x3CprojectId> "需求标题" "P1"

Step 2:进行需求拆解,生成任务文件

将需求文档按维度拆解,每个任务生成一个独立的 md 文件

  • 文件命名规范:{任务ID}-{任务名称}.md
  • 每个任务文件包含:任务概述、依赖、人天、相关上下文(数据库/接口/前端页面等)
  • 文件存放目录:documents/{项目名}/task_split/

Step 3:批量创建任务 + 附件挂载

⚠️ 关键原则:每个任务单独创建,附件挂载到对应任务,不要聚集上传。

批量创建任务的脚本示例:

tasks = [
    ('T01', '后端框架搭建', 'P0', 2, '基于clouddream-server初始化项目...'),
    ('T02', '数据库设计', 'P0', 3, '设计10张核心表结构...'),
    # ...更多任务
]

task_ids = {}  # 用于记录 taskId 映射

for task_id, name, priority, days, desc in tasks:
    title = f'{task_id}-{name}'
    # 1. 创建任务,获取 taskId
    result = call('create_task', {
        'requirementId': REQ_ID,
        'title': title,
        'priority': priority,
        'description': f'{desc} | {days}人天'
    })
    task_ids[task_id] = result['id']

    # 2. 等待一小段时间
    time.sleep(0.3)

# 3. 逐个上传附件到对应任务
for task_key, task_id in task_ids.items():
    filepath = f'documents/{项目名}/task_split/{task_key}-{任务名}.md'
    upload_file('task', task_id, filepath)  # entityType='task'!
    time.sleep(0.5)

Step 4:状态流转

# 任务开始时
python scripts/agentflow.py transition \x3CtaskId> pending in_progress "执行人"

# 任务完成时
python scripts/agentflow.py transition \x3CtaskId> in_progress completed "执行人"

# 需求全部完成
python scripts/agentflow.py transition \x3CrequirementId> in_progress completed "执行人"

常见错误避免

错误做法 正确做法
把所有任务文件上传到一个任务附件 每个任务单独创建+上传到自己的附件
entityType 填 requirement 上传任务附件 entityType 填 task 上传任务附件
上传到需求附件而不是任务附件 任务文档 → 任务附件;需求文档 → 需求附件

需求开发流程集成

每次接到需求文档时,必须自动同步到 AgentFlow:

阶段 AgentFlow 操作 状态值
开始拆解 create_project + create_requirement pending
逐步完成 create_task + transition in_progress
全部完成 requirement transition completed

详见 MEMORY.md「需求开发流程 × AgentFlow 集成规范」章节。

Usage Guidance
This skill appears to implement the AgentFlow API and file-upload features it advertises, but it contacts an unknown hard-coded IP (http://182.42.153.28:18900) and includes a convenient file-upload routine that will read and POST any local file you point it at. Before installing or using it, verify the endpoint's ownership and trustworthiness (ask the publisher for a homepage, documentation, or a signed release), confirm whether uploads require authentication or are logged/audited, and avoid pointing the upload commands at sensitive local files. If you cannot verify the server or need stronger controls, do not install or run the upload functionality; prefer an officially documented AgentFlow integration or require credentials and TLS-hostname verification.
Capability Analysis
Type: OpenClaw Skill Name: catfee-agent-flow Version: 1.9.0 The bundle provides a workflow management system that interacts with a remote MCP server at a hardcoded IP address (182.42.153.28). The most significant risk is the `upload_file` command in `scripts/agentflow.py` and the corresponding instructions in `SKILL.md`, which enable the agent to read local files and POST them to the remote server. While the stated purpose is to upload project-related 'task documents,' this functionality provides a direct mechanism for data exfiltration of any sensitive local file the agent can access.
Capability Assessment
Purpose & Capability
Name/description match the provided code and SKILL.md: the script implements project/requirement/task management and a multipart file upload flow. However, the tool always talks to a hard-coded external MCP endpoint (http://182.42.153.28:18900) with no documentation/homepage or authentication. Using a remote service is consistent with the stated purpose, but the lack of provenance for that host is unusual and deserves scrutiny.
Instruction Scope
SKILL.md explicitly instructs reading local files and uploading them (documents/{project}/...), and the script reads arbitrary filesystem paths and POSTs file contents to the remote endpoint. Reading and transmitting local files is within 'upload attachment' scope, but it also enables exfiltration of any local file if misused. The instructions encourage batch uploads and provide example code that iterates local directories — increasing the blast radius.
Install Mechanism
No install spec (instruction-only with a single Python script). This is lower-risk from an install/execution-injection perspective because nothing is downloaded at install time. The script does require a runtime with Python and the requests library, which is typical.
Credentials
The skill requests no environment variables or credentials. That could be reasonable for an internal service, but it also means the remote server accepts unauthenticated API calls (or relies on implicit session handling). Absence of required credentials is not inherently malicious, but when combined with an unknown external endpoint it increases risk: there is no account-level access control shown, and the server could receive arbitrary files/metadata.
Persistence & Privilege
The skill is not always-enabled and does not request system-level persistence or modification of other skills. It depends on the agent invoking the provided script; there are no elevated privileges requested in the package metadata.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install catfee-agent-flow
  3. After installation, invoke the skill by name or use /catfee-agent-flow
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.9.0
新增:需求全流程开发工作流文档(含任务创建+附件挂载规范);更新:附件上传 entityType 说明;更新:MEMORY.md 中 AgentFlow 集成规范
v1.8.0
新增:完整文件上传流程(upload_file命令);新增:get_upload_url工具;更新:完整工具列表和状态值说明
v1.0.1
修复:MCP地址改为环境变量读取,不再硬编码
v1.0.0
Initial release: AgentFlow 工作流系统 MCP 技能,支持项目管理、需求管理、任务管理、状态流转等13个工具
Metadata
Slug catfee-agent-flow
Version 1.9.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 4
Frequently Asked Questions

What is Catfee AgentFlow工作流?

AgentFlow 工作流管理系统 MCP 技能。当用户说"创建项目"、"添加需求"、"查看任务"、"更新状态"、"项目列表"、"需求列表"、"上传附件"等时触发。用于管理项目和任务的完整生命周期(创建→需求→任务→状态流转),以及为实体上传附件文件。 It is an AI Agent Skill for Claude Code / OpenClaw, with 132 downloads so far.

How do I install Catfee AgentFlow工作流?

Run "/install catfee-agent-flow" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Catfee AgentFlow工作流 free?

Yes, Catfee AgentFlow工作流 is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Catfee AgentFlow工作流 support?

Catfee AgentFlow工作流 is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Catfee AgentFlow工作流?

It is built and maintained by 冢猫 (@glory904649854); the current version is v1.9.0.

💬 Comments