Skill flagged — suspicious patterns detected

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

Weixin Plugin Installer

v0.1.3

通过聊天指令安装微信连接、刷新最新二维码、查询当前二维码状态、取消当前二维码刷新任务。

0· 107·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Pending
View report →
OpenClawOpenClaw
Suspicious
medium confidence
Purpose & Capability
The skill name/description (installing and managing Weixin QR/login) align with the scripts provided: installing/enabling an OpenClaw Weixin plugin, starting a login flow, rendering QR images from logs, querying status, cancelling tasks, and scheduling a gateway restart. Requested binaries (bash, python3, openclaw) are appropriate for these tasks.
Instruction Scope
SKILL.md narrowly restricts the agent to run only scripts under the skill directory, and the scripts themselves implement the described behaviors. One source-of-truth mismatch: SKILL.md's "绝对禁止" list includes package names like @tencent-weixin/openclaw-weixin, yet install_weixin_connection.sh internally calls 'openclaw plugins install "@tencent-weixin/openclaw-weixin"'. This is inconsistent (could be intentional to force use of the packaged script rather than ad-hoc commands) and should be clarified.
Install Mechanism
There is no separate install spec; the skill is instruction-plus-scripts (no arbitrary downloads embedded in the package). The only external install action is performed via the openclaw CLI (openclaw plugins install ...) which will fetch the plugin through OpenClaw's plugin mechanism — expected for a plugin installer.
!
Credentials
The skill requests no special environment variables, but scripts create state under $HOME/.openclaw/weixin-plugin-installer and a .out directory inside the skill, spawn background processes, and write logs that may contain sensitive tokens/cookies/session traces from 'openclaw channels login'. SKILL.md instructs not to expose tokens, but the scripts capture logs and may include tail snippets in emitted JSON messages. Ensure log file permissions and that emitting log tail won't leak credentials.
Persistence & Privilege
The skill creates persistent state (pid files, logs, cached QR images) in the user's home under ~/.openclaw and can schedule an asynchronous 'openclaw gateway restart' via a background job. That restart behavior is explained and intentionally deferred, but it is a privileged operation that can disrupt the gateway — acceptable for an installer but worth being aware of.
What to consider before installing
This skill appears to implement what it claims (installing/enabling a Weixin plugin and managing QR login), but there are a few things to consider before installing: - Clarify the contradiction in SKILL.md that "forbids" the plugin package names while the install script installs @tencent-weixin/openclaw-weixin; prefer to run only the provided scripts, not ad-hoc commands. - The scripts call 'openclaw plugins install' and 'openclaw channels login' which will fetch and run external plugin code via OpenClaw — only proceed if you trust the OpenClaw plugin source that will be installed. - Logs for the login flow are written under $HOME/.openclaw/weixin-plugin-installer; these logs may contain sensitive tokens/cookies/session traces. Check and secure filesystem permissions for that directory and review logs for sensitive strings before sharing. - The skill schedules an asynchronous 'openclaw gateway restart' which can interrupt service; run this in a maintenance window or on a host where a restart is acceptable. - Use only in an admin/private context as instructed. If you need higher assurance, request the exact plugin release/package to be installed (origin, checksum) and audit the plugin package fetched by openclaw before allowing the install. If you want, I can (a) list the exact file paths the skill will write to, (b) show the exact openclaw CLI commands it will run, or (c) suggest a minimal safe checklist to run before enabling this skill on production.

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

latestvk971azxs6q9gfaes2q09tvspjs83hysh

License

MIT-0
Free to use, modify, and redistribute. No attribution required.

Runtime requirements

💬 Clawdis
OSLinux · macOS
Binsbash, python3, openclaw

Comments