Back to skill

Security audit

微信连接

Security checks for vulnerabilities and agentic risk

Overview

This WeChat pairing skill appears intended for its stated purpose, but it needs Review because it can expose pairing tokens and local temporary files while making persistent install and configuration changes.

Install only if you are comfortable with this skill downloading a mutable npm installer, opening a local server, saving WeChat bot credentials, changing OpenClaw WeChat channel policy, and restarting the gateway. The current artifact should be reviewed or fixed before use because token exposure through `/tmp` and the local HTTP server could allow credential theft.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
Findings (5)

T09 · Insecure Skill Coding Practices

Error
Location
scripts/start.mjs:300
Finding

WeChat bot credentials exposed through an unauthenticated network service

Content
View full analysis
{ server.listen(HTTP_PORT, () => { log(`HTTP service started: http://localhost:${HTTP_PORT}`); resolve(server); }); }); ``` ```js case 'confirmed': updated.message = 'Login successful!'; updated.connected = true; updated.botToken = data.bot_token; updated.ilinkBotId = data.ilink_bot_id; updated.ilinkUserId = data.ilink_user_id; updated.baseUrl = data.baseurl; break; writeStatus(updated); ``` ### Technical Analysis The login-status endpoint serializes and returns the complete status object. After successful pairing, that object contains the WeChat bot token, bot identifier, user identifier, and API base URL. The HTTP server calls `server.listen(HTTP_PORT)` without specifying a loopback address. Node.js can therefore listen on all available interfaces rather than exclusively on `127.0.0.1`. The endpoint has no authentication or session authorization. It also sets `Access-Control-Allow-Origin: *`, allowing cross-origin browser code to read the response where browser private-network controls do not block the request. The browser UI only requires pairing state and a human-readable message. Returning the token and account identifiers exceeds the minimum data required by the declared functionality. ### Attack Path 1. A user starts the Skill, which opens TCP port 8765. 2. The user scans and confirms the WeChat QR code. 3. The WeChat API returns `bot_token`, `ilink_bot_id`, and `ilink_user_id`. 4. The Skill writes those values into the status object. 5. An attacker able to reach the ...[truncated 857 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
scripts/start.mjs:43
Finding

Authentication token stored in predictable files without restrictive permissions

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
scripts/start.mjs:448
Finding

Remote-derived user identifier interpolated into a shell command

Content
View full analysis
Remediation
View remediation

T03 · Remote Payload Retrieval and Execution

Warning
Location
scripts/start.mjs:480
Finding

Mutable remote npm package downloaded and executed automatically

Content
View full analysis
Remediation
View remediation

T05 · Unauthorized Access and Privilege Escalation

Warning
Location
scripts/start.mjs:326
Finding

HTTP server exposes unrelated files from the shared temporary directory

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (16)

Missing User Warnings

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill description and flow omit a clear upfront warning that it may auto-install a package, call external APIs, start a local server, open a browser, write persistent account data, modify messaging policy, and restart a background component. The embedded 'security explanation' also tells users to ignore warnings about dangerous code patterns, which is itself a red flag because it conditions users to bypass security signals.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
95% confidence
Finding

The skill performs network access to external WeChat APIs, starts a local HTTP server, opens a browser UI, and invokes installation/configuration actions, yet the manifest declares no tool scope or permissions. This deprives users and enforcement systems of accurate capability disclosure and makes the skill harder to review safely before execution.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill claims no command-line interaction is required, but it will automatically execute CLI installation and configuration commands behind the scenes. This is dangerous because it masks privileged system changes from the user and undermines informed consent about software installation and environment modification.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The trigger list includes an open-ended catch-all phrase, which increases the chance the skill activates on ambiguous user requests. Because this skill can install software, change configuration, write account data, and restart services, overly broad invocation raises the risk of unintended high-impact actions.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
97% confidence
Finding

Using npx -y @tencent-weixin/openclaw-weixin-cli@latest install pulls and executes the latest package version at runtime without pinning, creating a supply-chain execution risk. If the package is compromised or a breaking/malicious release is published, the skill will automatically run unreviewed code on the user's system.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

Beyond QR pairing, the skill writes persistent account files, updates global messaging policy, sets an allowlist, and triggers a background gateway restart. These are durable security-relevant state changes that extend far beyond a transient login flow and could alter message handling or trust boundaries without sufficient warning.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The manifest says the WeChat plugin install/connect flow requires no command-line interaction, implying a self-contained flow. However, the implementation depends on openclaw CLI checks, runs an npx installer, updates configuration via openclaw config set, and restarts the gateway via a spawned openclaw process, which is materially broader than the description suggests.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The script starts a local HTTP server and serves arbitrary files from /tmp, not just the specific QR assets needed for pairing. Any local website or process can query the open server, and exposing a broad /tmp file surface increases the risk of unintended data disclosure if sensitive temporary files exist or are planted there.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill does more than pairing: it persists bot credentials to disk and modifies channel policy/configuration, which exceeds the user's likely expectation from the description. Hidden persistence and policy changes are dangerous because they affect long-term trust boundaries and may silently expand or alter future message access.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The script writes the bot token and related account data to disk without explicit warning, consent, or file-permission hardening. Stored credentials can be stolen by other local users, malware, backups, or misconfigured sync tools, leading to account takeover or unauthorized bot access.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

Automatically installing software and restarting the gateway in the background without an explicit confirmation prompt removes an important user safety checkpoint. In a skill execution context, this can surprise users, disrupt service, or cause them to run unreviewed code and configuration changes they did not knowingly approve.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The script installs and executes @latest from npm via npx, which creates a supply-chain risk because the exact code fetched at runtime is not fixed or reviewable in advance. A compromised package, malicious maintainer update, or dependency hijack would run with the user's privileges during installation.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

This log line instructs the user to run an unpinned npx install command using @latest, encouraging execution of mutable remote code. Even though it is displayed as guidance, it still operationalizes a supply-chain exposure with no version review or reproducibility.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The error path again tells the user to execute an unpinned package installer from npm, perpetuating the same supply-chain risk outside the script's direct execution path. Because the skill is about installing an integration plugin that will later handle credentials, version drift is especially risky.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
93% confidence
Finding

Starting a local HTTP server and automatically opening a browser-based UI is behavior that should be disclosed in the manifest because it changes the local attack surface and user experience. While not inherently malicious, hidden service startup can surprise users and complicate security review, especially if the bind address or access controls are unspecified.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
95% confidence
Finding

The generated HTML is hardcoded with lang="zh-CN" and all user-facing instructions/messages are in Chinese. This imposes a language choice on users without opt-in or explanation, which matches the policy category for language or locale constraints.

Content

No source excerpt is available for this finding.