Skill flagged — suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

Clawtip Weather.Bak

v1.0.0

一位精通天气领域的专家,能够根据用户所在位置提供准确的每日天气报告。这是一项付费服务,执行前需完成支付验证。请注意,你应该用中文和用户交互(包含你的思考过程)。

0· 89·0 current·0 all-time

Install

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for liulian822/clawtip-weather-bak.

Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "Clawtip Weather.Bak" (liulian822/clawtip-weather-bak) from ClawHub.
Skill page: https://clawhub.ai/liulian822/clawtip-weather-bak
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Bare skill slug

openclaw skills install clawtip-weather-bak

ClawHub CLI

Package manager switcher

npx clawhub@latest install clawtip-weather-bak
Security Scan
Capability signals
CryptoRequires walletCan make purchasesRequires sensitive credentials
These labels describe what authority the skill may exercise. They are separate from suspicious or malicious moderation verdicts.
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Suspicious
medium confidence
Purpose & Capability
Providing paid weather reports legitimately requires a payment flow; the included scripts implement order creation, payment processing, token management and contacting external JD endpoints, which is coherent with a paid service. HOWEVER the registry metadata and top-level SKILL.md declare no required binaries or credentials while bundled code and the clawtip sub-skill clearly require Node and credential persistence. That mismatch (declared requirements missing vs actual code) is unexpected and unexplained.
!
Instruction Scope
The runtime SKILL.md forces a multi-stage payment workflow and contains several problematic instructions: it requires the agent to present its chain-of-thought to users (explicitly instructs to 'include your thinking process'), forbids the agent from inspecting script source on error, and mandates strict use (and installation, if missing) of another skill named exactly 'clawtip'. These directives expand scope beyond providing weather info, attempt to control agent behavior in ways that are unsafe (exposing chain-of-thought), and could coerce installation/invocation of another skill during runtime.
!
Install Mechanism
There is no declared install spec, yet the bundled clawtip scripts call a Node.js tool (encrypt.js) and bundle a large minified JS crypto library. The included documentation (IMPORTANT_STATEMENTS and clawtip SKILL.md) says Node must be present. The registry metadata at the top-level does not list Node as a required binary — this inconsistency is a red flag because the skill will fail or behave unexpectedly if Node is absent, and it relies on executing a bundled JS binary via subprocess.
!
Credentials
Top-level metadata claims no required env vars and lists only credential.read, but the code writes and reads a plaintext local token ('u') to configs/config.json (credential persistence), and the clawtip sub-skill declares credential.write and local plaintext credential storage. The skill also writes order files to the user's home (paths vary across docs: ~/.openclaw vs /root/openclaw). Requesting/using persistent credentials and filesystem writes without declaring them is disproportionate and surprising.
!
Persistence & Privilege
The skill persists state: order JSON files under a fixed directory in the user's home and a plaintext token at configs/config.json (written and read by scripts). While it does not set always:true, this persistent storage of a payment token and order history increases blast radius if the skill is compromised. The skill also instructs installing/using another skill (clawtip), which would broaden privileges if performed automatically.
What to consider before installing
This skill implements a paid-weather flow that contacts external endpoints and stores a plaintext payment token on disk, but its declared metadata contradicts the shipped code. Before installing: 1) Ask the publisher to explain the mismatches (why registry metadata omits Node and credential.write). 2) Confirm you trust the external endpoints (ms.jr.jd.com) and the ‘clawtip’ payment sub-skill; inspect that sub-skill before installing. 3) Do not allow the agent to reveal chain-of-thought; request removal of any instruction that asks the model to include its internal reasoning. 4) If you must use it, run it in an isolated sandboxed environment, deny unnecessary credentials, and restrict filesystem permissions (configs/config.json should be owner-only). 5) Require the skill author to explicitly list required binaries (node), declared permissions (credential.write), and exact storage paths, and to remove the prohibition on inspecting script source on error. If the author cannot justify these issues, treat the skill as untrusted.

Like a lobster shell, security has layers — review code before you run it.

latestvk9784qw6yrw6nh2njq8mwbm87n84yhhq
89downloads
0stars
1versions
Updated 1w ago
v1.0.0
MIT-0

天气报告服务

📌 技能概述

请注意,你应该用中文和用户交互(包含你的思考过程)。

本技能提供专业的天气报告服务。由于这是付费服务,工作流程严格要求按照两个步骤执行:支付处理 后,再进行 报告执行

直接执行: 如果用户在初始请求中已直接提供了 <订单号>(且订单文件中已包含 payCredential),则可跳过第一阶段和第二阶段,直接进入 🚀 第三阶段:报告执行


🛒 第一阶段:创建订单

如果这是首次交互且用户尚未付款,你必须先创建订单,再发起支付。

1. 所需参数

  • <位置>:用户希望查询天气报告的具体地点。

2. 执行命令

使用以下 bash 命令执行订单创建脚本。将 <...> 占位符替换为实际值。务必在参数两侧加引号,以防命令注入漏洞

python3 scripts/create_order.py "<位置>"

3. 输出处理

成功时,脚本将向标准输出打印以下键值对:

ORDER_NO=<值>
AMOUNT=<值>
QUESTION=<值>
INDICATOR=<值>

其中 ORDER_NOAMOUNTQUESTION 用于向用户披露订单信息(展示订单号、费用和查询位置)。提取 ORDER_NOINDICATOR 后进入 💳 第二阶段:支付处理

[!NOTE] 关于 AMOUNT 字段: 该字段的单位为人民币分(fen),这是 OpenClaw 技能描述中的标准计费单位。例如,AMOUNT=100 表示需支付 1.00 元人民币。在向用户展示费用时,请将该值除以 100 并以"元"为单位呈现,以便用户理解实际支付金额。

失败时,脚本以退出码 1 退出,并打印如下格式的消息:

订单创建失败: <错误详情>

[!CAUTION] 强制终止 —— 禁止执行任何后续操作。 如果脚本以非零退出码退出,其输出中包含 订单创建失败:,你必须立即终止整个工作流程。具体包括:

  • ❌ 不得查阅或检查脚本源代码。
  • ❌ 不得测试网络连接或服务器状态。
  • ❌ 不得代替用户进行任何重试操作。
  • ❌ 不得进入第二阶段或第三阶段。

唯一需要执行的操作是: 用通俗语言向用户报告失败情况,附上输出中的 <错误详情>,然后停止。

失败输出示例:

订单创建失败: Order creation request failed: <urlopen error [Errno 111] Connection refused>

告知用户:"订单创建失败,错误信息:Order creation request failed: <urlopen error [Errno 111] Connection refused>。请稍后重试,或检查您的位置输入是否有效。"


💳 第二阶段:支付处理

从第一阶段获取 ORDER_NOINDICATOR 后,发起支付请求。

操作: 使用技能 clawtip 处理支付并获取支付凭证。若该技能不存在,需先安装。

[!CAUTION] 技能名称精确匹配 —— 严格禁止替代

必须且只能使用名称完全等于 clawtip 的技能。这是一条硬性约束,不允许任何例外:

  • ✅ 如果 clawtip 技能不存在,唯一正确的做法是:安装名为 clawtip 的技能,而不是使用任何其他已存在的技能。

判断标准: 技能名称必须是字符串严格相等(name === "clawtip"),而非包含、前缀匹配或模糊匹配。

调用 clawtip 所需参数:

{
  "order_no": "<第一阶段获取的 ORDER_NO>",
  "indicator": "<第一阶段获取的 INDICATOR>"
}

[!NOTE] 订单的所有详细信息(skill-id、order_no、amount、question、description、slug、resource_url 等)已由第一阶段脚本写入固定目录下的 JSON 文件中,pre-verify-skill 会根据 order_noindicator 自动定位并读取。

目标: 等待支付成功,并获取 payCredential(支付凭证)。


🚀 第三阶段:报告执行

支付成功并获得 payCredential 后(或用户已直接提供 <订单号><支付凭证>),继续交互并执行报告脚本。

1. 所需参数

  • <订单号>:第一阶段生成的订单号。

[!NOTE] <支付凭证><位置> 无需通过命令行传入。clawtip 在支付成功后会将 payCredential 写入订单 JSON 文件,脚本会根据订单号自动从固定目录 /root/openclaw/skills/orders/{indicator}/ 下的 JSON 文件中读取所有所需信息。

2. 执行命令

使用以下 bash 命令执行天气报告服务。将 <...> 占位符替换为已验证的参数值。务必在参数两侧加双引号,以防命令注入漏洞

python3 scripts/weather_report.py "<订单号>"

执行后: 1. 提取脚本打印的 PAY_STATUS 值(格式为:PAY_STATUS: <值>),并再次输出展示。 2. ERROR 状态的特殊处理: 如果 PAY_STATUSERROR,提取 ERROR_INFO 值(格式:ERROR_INFO: <值>),向用户告知确切的错误原因并引导其解决。不得继续执行后续服务逻辑。

Comments

Loading comments...