功能描述

Use for JumpServer V4 preflight, `.env.local` initialization, org selection, and read-only asset, permission, audit, and access queries through the bundled `...

使用说明 (SKILL.md)

JumpServer Skills

Name: JumpServer
Author: liqiang-fit2cloud

JumpServer 查询型 skill：允许环境初始化写入（.env.local 与 JMS_ORG_ID），但不执行业务对象或权限的创建、更新、删除、追加或移除。

Input / Output / 输入输出

类型	常见输入	返回
环境初始化	`JMS_API_URL`、鉴权信息、可选 `org-id`	配置完整性、`.env.local` 写入结果、连通性、组织写入结果
对象查询	`resource`、`id`、`name`、`filters`	列表、详情、解析结果
权限与审计	`filters`、`audit-type`、时间范围、`command_storage_id`	权限详情、审计详情、最近活动
访问分析	`username`、可选 `asset-name`	用户可访问资产/节点、单资产访问视图

Route / 路由流程

config-status --json
  -> complete=false ? collect env info -> config-write --confirm
  -> ping
  -> org missing/inaccessible ? select-org [--org-id] -> select-org --confirm
  -> read-only query

仅当可访问组织集合恰好是 {0002} 或 {0002,0004} 时，才允许自动写入 0002。

Capability Matrix / 能力矩阵

Intent	Must Use	Precheck	Output	Stop If
初始化环境	`jms_diagnose.py config-status/config-write/ping/select-org`	无	配置状态、`.env.local` 写入结果、连通性、组织持久化结果	地址或鉴权缺失、地址不可达、组织不可访问
查资产与对象	`jms_assets.py list/get`、`jms_diagnose.py resolve/resolve-platform`	`config-status --json` -> 必要时 `config-write --confirm` -> `ping` -> 必要时 `select-org --confirm`	资产类列表、详情、对象解析结果	名称不唯一、对象不清楚、组织未准备好
查权限规则	`jms_permissions.py list/get`	`config-status --json` -> 必要时 `config-write --confirm` -> `ping` -> 必要时 `select-org --confirm`	权限列表、权限详情	组织未准备好
查审计记录	`jms_audit.py list/get`	`config-status --json` -> 必要时 `config-write --confirm` -> `ping` -> 必要时 `select-org --confirm`	登录、操作、会话、命令审计	`audit-type=command` 缺 `command_storage_id`
做访问分析	`jms_diagnose.py user-assets/user-nodes/user-asset-access/recent-audit`	`config-status --json` -> 必要时 `config-write --confirm` -> `ping` -> 必要时 `select-org --confirm`	用户可访问资产/节点、单资产访问、最近审计	用户不存在、候选过多、组织未准备好

Core Rules / 核心规则

Rule	Required Behavior
预检顺序	`config-status --json -> config-write --confirm（如需） -> ping -> select-org --confirm（如需） -> read-only query`
环境写入	允许通过 `config-write --confirm` 生成或更新 `.env.local`
组织写入	允许通过 `select-org --confirm` 持久化 `JMS_ORG_ID`
保留组织特判	可访问组织集合仅在 `{0002}` 或 `{0002,0004}` 时才自动写入 `0002`
范围边界	把这个 skill 当作查询型 skill，而不是通用运维执行器
审计默认窗口	`date_from/date_to` 省略时默认最近 7 天
命令审计	`audit-type=command` 时必须提供 `command_storage_id`
非支持动作	遇到 create/update/delete/append/remove/unblock 时直接说明“业务动作只保留查询”

Canonical Commands / 命令骨架

配置检查与写入：

python3 scripts/jms_diagnose.py config-status --json
python3 scripts/jms_diagnose.py config-write --payload '{"JMS_API_URL":"https://jump.example.com","JMS_ACCESS_KEY_ID":"\x3Cak>","JMS_ACCESS_KEY_SECRET":"\x3Csk>","JMS_VERSION":"4"}' --confirm
python3 scripts/jms_diagnose.py ping

组织选择：

python3 scripts/jms_diagnose.py select-org
python3 scripts/jms_diagnose.py select-org --org-id \x3Corg-id>
python3 scripts/jms_diagnose.py select-org --org-id \x3Corg-id> --confirm

对象查询：

python3 scripts/jms_assets.py list --resource user --filters '{"username":"openclaw"}'
python3 scripts/jms_assets.py get --resource asset --id \x3Casset-id>
python3 scripts/jms_diagnose.py resolve --resource node --name demo-node
python3 scripts/jms_diagnose.py resolve-platform --value Linux

权限与审计查询：

python3 scripts/jms_permissions.py list --filters '{"limit":20}'
python3 scripts/jms_permissions.py get --id \x3Cpermission-id>
python3 scripts/jms_audit.py list --audit-type operate --filters '{"limit":30}'
python3 scripts/jms_audit.py get --audit-type command --id \x3Ccommand-id> --filters '{"command_storage_id":"\x3Ccommand-storage-id>"}'

Success Criteria / 成功标准

先完成或明确阻塞在 config-status -> ping -> select-org 预检链路上，不跳步。
配置缺失时，能按字段收集用户回复并在确认后调用 config-write --confirm。
组织缺失时，能先返回候选组织，再在确认后调用 select-org --confirm。
查询请求走正式 jms_*.py 入口，返回结果或明确说明阻塞原因。
对业务写操作直接拒绝，不绕过正式入口另写临时 SDK/HTTP 脚本。

Not For / 不适用

不适用于资产、平台、节点、账号、用户、用户组、组织、权限的创建、更新、删除。
不适用于追加关系、移除关系、解锁用户、改密、批量修改。
不适用于临时 SDK/HTTP 脚本绕过正式入口。

安全使用建议

This repository appears to be a legitimate JumpServer read-only helper, but there are several things to consider before installing: (1) the registry metadata declares no required env vars, yet the code will ask for JMS_API_URL and credentials (AK/SK or username/password) and will write them to a local .env.local — be comfortable with storing those secrets on disk in the skill directory. (2) The runtime disables HTTPS certificate verification (verify=False), increasing MITM risk; avoid using production credentials against untrusted networks or ask the maintainer to make verification configurable. (3) The bootstrap can auto-install the jumpserver SDK via pip if you allow it — review and trust the PyPI package/version before confirming installs. (4) The skill enforces read-only query behavior in code, but the agent will persist JMS_ORG_ID automatically in a special-case; review that behavior if you care about persisted organization selection. Recommended actions: inspect the scripts yourself, run them in an isolated environment or container, limit which JumpServer account/credentials you provide (use a least-privilege read-only account), and ensure .env.local is secured or not used for sensitive production credentials.

功能分析

Type: OpenClaw Skill Name: jumpserver-skills Version: 1.0.0 The skill bundle provides administrative query and configuration capabilities for JumpServer V4. It contains several high-risk behaviors: `jms_diagnose.py` allows the agent to persist sensitive credentials (AK/SK and passwords) to a local `.env.local` file, and `jms_bootstrap.py` performs automated dependency installation using `pip`. Additionally, `jms_runtime.py` explicitly disables SSL certificate verification (`verify=False`), which is a significant security vulnerability (MITM risk). While these actions are aligned with the stated purpose of environment initialization, the combination of secret persistence, automated execution, and weakened transport security warrants a suspicious classification.

能力评估

⚠ Purpose & Capability

The skill's name/description say it's a read-only JumpServer preflight/query helper, which matches the included scripts' behavior. However the registry metadata declares no required environment variables or primary credential while the implementation plainly expects JMS_API_URL and either AK/SK or username/password (and will persist them). The manifest omission is an incoherence: a JumpServer integration legitimately needs those env vars, so they should be declared.

⚠ Instruction Scope

SKILL.md instructs the agent to run local CLI scripts that will: collect credentials interactively, call the SDK against the JumpServer API, and write a .env.local file (including secrets) via config-write and select-org --confirm. The skill claims 'read-only' for business objects, and the code enforces query-only behavior, but the instructions and code explicitly allow writing .env.local and persisting JMS_ORG_ID (including an automatic special-case write for certain org sets). That means the skill will collect and persist sensitive credentials and perform network calls — this scope is broader than the registry metadata indicates.

ℹ Install Mechanism

There is no registry install spec, but the code includes a bootstrap helper that will detect missing Python packages from requirements.txt and (if invoked with the confirm flag) call pip to install them. The requirements file only lists jumpserver-sdk-python, and the bootstrap enforces simple requirements, so the install path is typical for Python CLI tools. Automatic pip install via subprocess is a moderate-risk action (it writes code to disk from PyPI), but it is expected for a local CLI script.

⚠ Credentials

The implementation requires JMS_API_URL and either JMS_ACCESS_KEY_ID/JMS_ACCESS_KEY_SECRET or JMS_USERNAME/JMS_PASSWORD and will write them into .env.local. None of these required credentials were declared in the skill registry metadata. Additionally, scripts disable HTTPS verification (verify=False) and suppress certificate warnings, which reduces transport security and increases risk if credentials are used against an intercepted connection.

ℹ Persistence & Privilege

The skill writes persistent configuration (.env.local) and JMS_ORG_ID into its own skill directory and reloads that config. It does not set always:true and does not modify other skills' configs. Persisting secrets to a local file is expected for this tool but is a privileged action (persisting credentials to disk); users should ensure the file's location and filesystem permissions are acceptable.

版本历史

v1.0.0

Initial release of jumpserver-skills - Provides read-only querying and environment initialization for JumpServer V4 via bundled CLI scripts. - Supports `.env.local` generation/updating and persistent org selection (`JMS_ORG_ID`). - Allows listing and retrieving assets, users, nodes, permissions, and audit details. - Blocks all create, update, delete, append, or remove actions for business objects or permissions. - Preflight checks enforce a strict flow: config status → config write → ping → org selection → query. - Only executes queries through official CLI entry points; temporary SDK/HTTP scripts not supported.

元数据

Slug jumpserver-skills

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 1

常见问题

JumpServer 是什么？

Use for JumpServer V4 preflight, `.env.local` initialization, org selection, and read-only asset, permission, audit, and access queries through the bundled `... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 112 次。

如何安装 JumpServer？

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

JumpServer 是免费的吗？

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

JumpServer 支持哪些平台？

JumpServer 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 JumpServer？

由 liqiang-fit2cloud（@liqiang-fit2cloud）开发并维护，当前版本 v1.0.0。

JumpServer