Back to skill

Security audit

Aiding Shifu

Security checks for vulnerabilities and agentic risk

Overview

This skill matches its stated Aiding Shifu login and work-order purpose, but it handles account tokens and personal data in unsafe, under-disclosed ways that need review before installation.

Do not install this version unless you are prepared to review and contain credential exposure. The publisher should remove the packaged config.json secret, revoke the exposed token, stop sending login payloads to third-party QR services, avoid logging or displaying raw tokens, replace wildcard postMessage token transfer, and store credentials in a protected secret store with clear user consent.

Vulnerability Patterns
  • 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
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (6)

T09 · Insecure Skill Coding Practices

Error
Location
tools/login/login.js:126
Finding

Authentication QR Payload Disclosed to an Unrelated Third-Party Service

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
config.json:1
Finding

Bearer Token and Personal Phone Number Shipped in Plaintext Configuration

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
tools/login/qr-login.html:345
Finding

Bearer Tokens and User Information Exposed Through Wildcard Cross-Window Messaging

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
tools/login/sms-login.html:397
Finding

Bearer Tokens Persisted in Script-Readable Plaintext Storage

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
tools/login/sms-login.html:310
Finding

Phone Number and One-Time Verification Code Transmitted in Request URLs

Content
View full analysis
1. `GET /auth/sendVerifyCode?phoneNumber={手机号}` - 发送验证码 2. `POST /auth/sms/login?phoneNumber={手机号}&code={验证码}` - 验证码登录(URL参数) ``` ### Technical Analysis The SMS login endpoint places both the phone number and one-time verification code in the query string. Changing the HTTP method to POST does not protect parameters that remain in the URL. HTTPS protects the request from ordinary network interception while in transit, but URLs are frequently recorded by reverse proxies, API gateways, server access logs, monitoring products, browser tooling, debugging systems, and error reports. The send-code endpoint similarly exposes the phone number in a URL. The OTP is an authentication secret during its validity period. Logging it creates a secondary credential-distribution path unrelated to the minimum requirements of SMS authentication. ### Attack Path 1. The victim submits a phone number and valid SMS verification code. 2. The browser constructs a URL containing both values. 3. A reverse proxy, API gateway, server logger, monitoring agent, or debugging system records the request URL. 4. A party with access to those records extracts the phone number and OTP before it expires. 5. The attacker submits the credentials to the login endpoint and obtains an account tok ...[truncated 393 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
tools/login/run-login.js:51
Finding

Authentication Tokens and Raw Login Responses Written to Console Output

Content
View full analysis
{ console.log('\n返回数据:', r); ``` ### Technical Analysis `run-login.js` prints the complete token returned by the login API. `login.js` prints the first 50 characters, which may reveal substantial JWT metadata and sensitive credential material even though it is truncated. The test utility prints complete authentication responses and uses a hard-coded scene identifier, so any token or personal data included in the response is exposed to console output. Console output is commonly retained in terminal scrollback, CI logs, process supervisors, support bundles, remote session recordings, and centralized log services. Authentication credentials should never be treated as ordinary diagnostic data. ### Attack Path 1. A login or status-check script executes in a shared terminal, CI job, managed process, or recorded support session. 2. The API returns a token or authentication response. 3. The script writes the credential or raw response to standard output. 4. Console output is retained or forwarded to a log system. 5. A user with log access extracts and replays the token before it expires. ### Impact Assessment Disclosure of a complete bearer token allows account impersonation for its validity period. Raw login responses may also reveal personal information, identifiers, or ...[truncated 115 chars]
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (26)

Tp4

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The skill metadata and documentation describe a broad platform-management skill, but the analyzed behavior includes risky login-centric actions such as fetching QR content from a third-party service, invoking a local application via exec, and exposing authentication artifacts to console output. This mismatch is dangerous because users and reviewers may grant broad trust and permissions based on the stated purpose while the implemented behavior handles credentials and local execution in ways not clearly disclosed.

Content

No source excerpt is available for this finding.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 95)May include surrounding context.

md
**工具路径**:`tools/login/sms-login.html`

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 111)May include surrounding context.

md
**工具路径**:`tools/login/qr-login.html`

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

This login page explicitly exposes the token to the user interface, clipboard, local storage, and window.opener without any warning or minimization. Authentication tokens are bearer secrets; any unnecessary exposure enables account takeover if another script, window, extension, or local user obtains the value.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The script prints authentication material such as a token or identifier directly to stdout after login. Console output is often captured in terminal history, logs, CI systems, or screen recordings, which can enable session hijacking or unauthorized account access if the token is reusable.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

Printing the login token without masking or warning exposes sensitive credentials to anyone with access to the terminal session or its logs. Because this is a login helper, the exposed value is especially sensitive and may grant direct authenticated access to the platform.

Content

No source excerpt is available for this finding.

Hidden Instructions

High
Category
Prompt Injection
Confidence
70% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · tools/login/sms-login.html (reported line 187)May include surrounding context.

html
<div class="logo">艾登<span>师傅</span></div>
        <div class="subtitle">短信验证码登录 WorkBuddy</div>
        
        <!-- 登录表单 -->
        <div id="loginForm">
            <div class="form-group">
                <label>手机号</label>

Missing User Warnings

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The page exposes the authentication token directly in the UI, copies it to the clipboard, stores it in localStorage, and sends it to window.opener using a wildcard target origin. This creates multiple easy exfiltration paths for bearer credentials via XSS, malicious parent/opener windows, shoulder-surfing, browser extensions, or shared-device access; in this skill context, the page is explicitly for platform login, so token theft would directly enable account/session compromise.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The skill explicitly states that access tokens are saved to a local config.json path but does not present this as a user-facing security warning or explain retention, access controls, or risks. Local token storage can enable session theft or unauthorized reuse if the file is readable by other tools, users, or processes on the host.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill processes sensitive personal data during login and work-order creation, including phone numbers, customer names, addresses, and potentially technician contact details, but does not clearly warn users about transmission and handling of this data. In this business context, undisclosed collection and transfer of personal information increases privacy, compliance, and misuse risk.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

Broad trigger phrases for login can cause the skill to activate unexpectedly during ordinary conversation, increasing the chance that users are funneled into an authentication workflow without clear intent. In a skill that stores tokens and transmits login data, unintended activation raises the risk of accidental credential handling and privacy-impacting actions.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The script sends the full login QR payload URL to a third-party QR generation service, exposing authentication material outside the intended platform. Anyone with access to that URL content may be able to render the QR code and potentially complete or interfere with the login flow, and it also leaks usage metadata to an unrelated external service.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
96% confidence
Finding

This external transmission sends login-related QR content to https://api.qrserver.com/, a third party unrelated to the primary authentication service. Because the QR encodes the login URL/ticket, the transmission increases exposure of sensitive authentication state and creates unnecessary dependency on an outside service during login.

Content

Scanner excerpt · tools/login/login.js (reported line 128)May include surrounding context.

js
// 生成二维码图片(使用在线 API)
function generateQrCodeUrl(text) {
    // 使用腾讯的 API 生成二维码
    return `https://api.qrserver.com/v1/create-qr-code/?size=200x200&data=${encodeURIComponent(text)}`;
}

// 检查登录状态

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The skill is described as helping users log in to the 艾登师傅 platform and manage installation-service workflows. While opening a QR image is related to login, this implementation uses OS shell commands through exec() to launch arbitrary paths/URLs, which introduces a general local command-execution capability not inherently required by the skill's stated business purpose.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The script stores the access token and associated user data in plaintext under the user's home directory without warning, consent, or permission hardening. Local malware, other users on a shared system, backups, or accidental disclosure could expose the token and allow unauthorized access to the account or APIs.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The page handles the authentication token in multiple unsafe ways: it displays the token in the UI, stores it in localStorage, and propagates it to another window. For a login flow, this greatly expands the token exposure surface, making theft via XSS, malicious parent/opener windows, shared-device access, or browser extensions much easier.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
97% confidence
Finding

The page sends authentication-flow data to https://api.qrserver.com/ to create the QR image. In a login skill, external transmission of login URLs is more dangerous because it leaks sensitive workflow data to an unnecessary third party and can aid replay, tracking, or abuse of the login session.

Content

Scanner excerpt · tools/login/qr-login.html (reported line 214)May include surrounding context.

html
// 生成二维码图片 URL
        function generateQrCodeUrl(text) {
            return `https://api.qrserver.com/v1/create-qr-code/?size=200x200&data=${encodeURIComponent(text)}`;
        }

        async function createQRCode() {

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The QR code is generated by embedding the login URL into a third-party QR service request, which transmits the login URL to an external domain. Even if the URL is not itself a final token, it is clearly part of the authentication flow and may contain identifiers or one-time login material that should not be disclosed to unrelated third parties.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

Multiple user-facing strings are hard-coded in Chinese, and the script does not offer an alternate locale or let the user choose their preferred language. This is a natural-language policy issue because it imposes a specific language on all users without documented justification or opt-in.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
95% confidence
Finding

The script sends the login URL returned by the platform to a third-party QR generation service, exposing authentication-related login state to an external domain outside the primary service. Even if the URL is intended for scanning, sharing it with a separate provider increases privacy and account security risk because that provider can observe or log the login URL and associated metadata.

Content

Scanner excerpt · tools/login/run-login.js (reported line 22)May include surrounding context.

js
console.log('SceneId:', data.data.sceneId);
        console.log('请用微信扫码...\n');
        
        const qrUrl = 'https://api.qrserver.com/v1/create-qr-code/?size=300x300&data=' + encodeURIComponent(data.data.url);
        const qrPath = path.join(os.tmpdir(), 'aiding-login.png');
        
        https.get(qrUrl, (r) => {

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
86% confidence
Finding

The manifest scope covers aiding-shifu platform login and work-order related tasks, and network access to obtain a login QR code is expected. However, spawning a child process via exec to launch a local file viewer is a broader host-execution capability that is not inherent to that purpose and materially expands what the script can do on the user's machine.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The document language is set to zh-CN and all visible interface text is presented only in Chinese, with no indication that the user can choose another language. This is a natural-language policy concern because it imposes a specific language/locale without opt-in or documented justification.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

This code makes an HTTPS request to an external server using a login status endpoint and includes the sceneId in the request path. Although it logs the returned data, there is no user-facing disclosure or confirmation before transmitting this login-related value off-host, which is a safety-relevant network operation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
95% confidence
Finding

Natural-language strings in the file, including the title, usage context, status messages, and errors, are presented only in Chinese. This forces a specific language on users without opt-in or a documented justification for the locale constraint.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
97% confidence
Finding

The document is explicitly marked as zh-CN and all user-facing text is in Chinese, with no indication that users can select another language. Under the stated policy, forcing a specific language without opt-in is a natural-language policy concern unless the locale limitation is clearly justified.

Content

No source excerpt is available for this finding.