← 返回 Skills 市场
shawnhan98

Feishu Task Workbench

作者 shawnhan98 · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
150
总下载
0
收藏
0
当前安装
4
版本数
在 OpenClaw 中安装
/install feishu-task-workbench
功能描述
在单个 Feishu/OpenClaw 对话窗口中运行多任务工作台,让用户在一个窗口里完成任务新建、切换、继续、总结、关闭与归档,并通过 `sessions_spawn`、`sessions_send`、`sessions_history` 把每个任务路由到独立会话。**当用户发送“任务列表 / 新建任务 / 切到
使用说明 (SKILL.md)

飞书任务工作台

在一个 Feishu/OpenClaw 对话窗口中运行单入口、多任务工作台。

对用户只暴露任务概念:新建列表切换继续总结关闭归档

任务状态以 scripts/task_registry.py 为唯一事实来源。

触发信号(强匹配)

当用户出现以下任一表达时,应优先激活本技能,而不是进入泛化问答澄清:

  • 任务列表
  • 新建任务:...
  • 切到 #\x3Cid> / 切换到 #\x3Cid>
  • 继续(在任务上下文中)
  • 总结 #\x3Cid>
  • 关闭 #\x3Cid>
  • 归档 #\x3Cid>
  • 任务状态

同义英文命令(task listtask newtask use 等)同样适用。

强制能力门禁(必需)

如果缺少以下任一会话工具,本技能必须阻断执行

  • sessions_spawn
  • sessions_send
  • sessions_history

执行规则:

  1. 在会话中的第一次任务命令时检查能力。
  2. 如果缺任何工具,立即停止并返回阻断提示。
  3. 阻断状态下不允许创建或更新任务 registry。
  4. 明确提示如何修复主机配置(tools.allowagentToAgentsessions.visibility)。

禁止静默降级为 registry-only 模式。

主机配置提醒

要获得真实任务隔离,通常需要:

  • agent allowlist 包含 sessions_spawnsessions_sendsessions_history,通常还需要 sessions_list
  • tools.agentToAgent.enabled=true
  • tools.sessions.visibility=all

参考配置:

{
  "tools": {
    "agentToAgent": { "enabled": true },
    "sessions": { "visibility": "all" }
  },
  "agents": {
    "list": [
      {
        "id": "main",
        "tools": {
          "allow": [
            "sessions_spawn",
            "sessions_send",
            "sessions_history",
            "sessions_list"
          ]
        }
      }
    ]
  }
}

核心运行模型

  • 一个用户可见的飞书聊天窗口
  • 每个飞书账号+联系人对应一个任务 registry 文件
  • 一个当前任务指针
  • 每个任务对应一个独立会话
  • 非全局命令默认路由到当前任务

除非用户明确要求查看实现细节,否则不要暴露 sessionKey

Registry 路径策略

使用“联系人隔离”路径:

tasks/feishu/\x3Caccount>/\x3Cpeer>.json

规则:

  • \x3Caccount> 使用当前飞书账号 id
  • \x3Cpeer> 使用当前飞书用户 id(例如 ou_xxx),并做文件名安全化
  • 不同 account+peer 严禁共享同一 registry 文件
  • Feishu 与 Weixin 使用不同根目录,严禁共用同一 registry

首次使用先初始化:

python3 scripts/task_registry.py --registry tasks/feishu/\x3Caccount>/\x3Cpeer>.json init

标准命令

  • 新建任务:\x3Ctitle>
  • 任务列表
  • 切到 #\x3Cid>
  • 继续
  • 总结 #\x3Cid>
  • 关闭 #\x3Cid>
  • 归档 #\x3Cid>
  • 任务状态

可兼容英文等价命令(如 task newtask listtask use)。

控制面动作

1. 新建任务

  1. 先执行能力门禁(仅首次任务命令)
  2. sessions_spawn 创建任务专属会话
  3. 用 registry add --session-key ... --make-current 持久化任务
  4. 回执任务 id 与当前任务

2. 查看任务列表

执行:

python3 scripts/task_registry.py --registry \x3Cpath> list

返回精简列表,并标记当前任务。

3. 切换任务

执行:

python3 scripts/task_registry.py --registry \x3Cpath> switch \x3Cid>

回执已切换目标。

4. 继续当前任务

  1. 读取当前任务(show
  2. 使用 sessions_send 把消息路由到当前任务 sessionKey
  3. 需要时将状态更新为 in_progress

若无当前任务,提示先新建或先查看列表。

5. 总结任务

  1. 读取 registry(summarize \x3Cid>
  2. 需要时调用 sessions_history 补充上下文
  3. 输出“进展/产出/阻塞/下一步”
  4. 可选写回 update --summary

6. 关闭 / 归档任务

执行:

python3 scripts/task_registry.py --registry \x3Cpath> close \x3Cid> --summary "..."
python3 scripts/task_registry.py --registry \x3Cpath> archive \x3Cid> --summary "..."

7. 任务状态(替代 /status 的任务视图)

当用户发送 任务状态 时:

  1. 读取 show(当前任务)+ list(任务清单)
  2. 返回当前任务编号、标题、状态
  3. 返回当前任务绑定的 sessionKey(仅用于排障)
  4. 返回 registry 路径(便于定位持久化文件)

说明:/status 是 OpenClaw 全局会话状态,不等价于任务工作台的任务路由状态。

回复样式约束

对于路由到具体任务的非全局回复,必须加任务头:

[任务:#2 周报]

该任务头需要在每次任务回复中可见,便于用户核对路由是否正确。

安全使用建议
Before installing, consider these points: - The skill will create and modify JSON files under tasks/feishu/<account>/<peer>.json on the agent host. Confirm the host working directory and file permissions are safe and that the agent runs with least privilege (or run the skill inside a sandbox/container). - The included Python script does not sanitize or validate the registry path components. Ensure the agent or caller constructs safe account/peer values (no ../ or absolute paths). Ask the author to add filename sanitization (e.g., whitelist/escape peer ids) or patch scripts/task_registry.py to reject/escape unsafe path segments. - The skill stores sessionKey values in the registry and the instructions may return sessionKey and the registry path in status responses. Treat sessionKey as sensitive: verify you are comfortable with that exposure, or request the skill only show sessionKey on explicit explicit user request and never include it in normal status outputs. - The skill requires the platform session tools (sessions_spawn, sessions_send, sessions_history). Enabling agent-to-agent/session visibility increases the blast radius: only grant those capabilities to trusted skills and consider limiting which agents/skills can use them. - If you need higher assurance, run the skill in a controlled environment first (test account or container), review/modify scripts/task_registry.py to add path sanitization and stricter file permissions, and verify the skill's runtime behavior (that it won't write outside tasks/feishu or leak session identifiers).
功能分析
Type: OpenClaw Skill Name: feishu-task-workbench Version: 1.0.0 The skill implements a multi-tasking workbench for Feishu/OpenClaw, using a local JSON registry to track tasks and routing messages to independent sessions via standard OpenClaw session tools. The Python script `scripts/task_registry.py` is a straightforward CRUD utility for task management, and the instructions in `SKILL.md` focus on task orchestration, session isolation, and capability checks without any evidence of malicious intent, data exfiltration, or unauthorized execution.
能力评估
Purpose & Capability
Name/description ask for per-task sessions and a registry; the SKILL.md and included scripts only require session tooling (sessions_spawn/sessions_send/sessions_history) and a local JSON registry. No unrelated cloud credentials or external binaries are requested — the requested capabilities are coherent with the stated purpose.
Instruction Scope
Instructions correctly constrain behavior to session routing and registry operations, but they (1) instruct the skill to return sessionKey and registry path in the '任务状态' view (exposing internal identifiers), and (2) include inconsistent language about not exposing sessionKey vs. 'include sessionKey for debugging'. The runtime guidance does not require or instruct explicit sanitization of account/peer components used to build file paths, creating a potential path traversal/file-overwrite risk.
Install Mechanism
Instruction-only skill with a small included Python script; there is no external install/download step or remote code fetch. Risk from install mechanism is low.
Credentials
The skill requests no environment variables or credentials. It does persist and read sessionKey values returned by sessions_spawn, which is expected for routing but is sensitive. Storing sessionKey in local files is functional but raises privacy/secret-management considerations.
Persistence & Privilege
The skill writes registry JSON files under tasks/feishu/<account>/<peer>.json in the host filesystem (created by scripts/task_registry.py). Writing persistent files is expected, but because the script uses the provided path without programmatic sanitization, a crafted account/peer value could cause files to be created/overwritten outside the intended directory. The skill does not request elevated platform privileges or always:true.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install feishu-task-workbench
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /feishu-task-workbench 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.0
Align with stable weixin-task-workbench protocol: enforce sessions gate, per-channel isolated registry, task-status command, task-header replies, and bundled deterministic registry script.
v0.1.2
Add important host-configuration reminders and a runtime self-check checklist. The skill now documents the OpenClaw policies typically required for sessions_spawn/sessions_send/sessions_history, clarifies full-workbench vs registry-only mode, and tells agents to disclose degraded mode early instead of implying real session isolation.
v0.1.1
Clarify host capability requirements and runtime modes. The skill now documents its dependency on sessions_spawn, sessions_send, and sessions_history, explains honest registry-only degradation in restricted hosts, and defines an end-to-end acceptance standard for true multi-task isolation.
v0.1.0
Initial public release: turn a single Feishu/OpenClaw chat into a multi-task workbench. Create, switch, continue, summarize, and close parallel workstreams with a local JSON registry and deterministic task lifecycle helper script.
元数据
Slug feishu-task-workbench
版本 1.0.0
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 4
常见问题

Feishu Task Workbench 是什么?

在单个 Feishu/OpenClaw 对话窗口中运行多任务工作台,让用户在一个窗口里完成任务新建、切换、继续、总结、关闭与归档,并通过 `sessions_spawn`、`sessions_send`、`sessions_history` 把每个任务路由到独立会话。**当用户发送“任务列表 / 新建任务 / 切到. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 150 次。

如何安装 Feishu Task Workbench?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install feishu-task-workbench」即可一键安装,无需额外配置。

Feishu Task Workbench 是免费的吗?

是的,Feishu Task Workbench 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Feishu Task Workbench 支持哪些平台?

Feishu Task Workbench 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Feishu Task Workbench?

由 shawnhan98(@shawnhan98)开发并维护,当前版本 v1.0.0。

💬 留言讨论