Back to skill

Security audit

product-doc-hub

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a product-document template, but its API console assets include under-scoped credential, telemetry, and live-API patterns that should be reviewed before use.

Review this before installing or reusing it. Treat the API page as a mockup, not a safe production console: remove or mask demo keys, remove credential and telemetry endpoints from defaults, add sandbox-only guidance and confirmations for mutating requests, and fix the DOM XSS/key-generation issues before hosting it on a real origin.

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 (3)

T09 · Insecure Skill Coding Practices

Error
Location
assets/api.html:1294
Finding

DOM-Based Cross-Site Scripting Through Unsanitized API Key Name

Content
View full analysis
' + '
' + name + ' 活跃
' + newKey + '
' + '
0
本月调用
今天
创建日期
' + '
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
assets/api.html:1301
Finding

API-Key-Like Credentials Generated With a Non-Cryptographic PRNG

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Note
Location
assets/prd.html:64
Finding

External Link Opened Without Opener Isolation

Content
View full analysis
在线体验 ``` ### Technical Analysis The link opens a cross-origin destination using `target="_blank"` but does not explicitly specify `rel="noopener noreferrer"`. In browser environments that retain an opener reference for such links, the destination can access `window.opener` and navigate the original tab. Modern browsers often apply implicit `noopener` behavior to `_blank` links, reducing practical exploitability. Explicit isolation remains necessary for consistent behavior across browser versions, embedded webviews, and unusual deployment environments. ### Attack Path 1. A user opens the PRD page and follows the external online-experience link. 2. The destination opens in a new tab. 3. If the browser or webview exposes `window.opener`, the destination obtains a reference to the original PRD tab. 4. A compromised or malicious destination executes code such as: ```javascript window.opener.location = 'https://attacker.example/phishing'; ``` 5. The original tab is replaced with an attacker-controlled page. 6. The user may trust the replacement because it appeared in a tab originating from the legitimate document. ### Impact Assessment The destination may be able to redirect or replace the original tab, enabling phishing, brand impersonation, or deceptive credential prompts. This issue does not grant direct access to the original page's DOM when normal cross-origin restrictions apply, and it does not independently provide system-level privileges. ]]>
Remediation
View remediation
Online Experience ``` Apply the same rule to every external link that uses `target="_blank"`. Where opening a new tab is unnecessary, remove `target="_blank"` entirely. Consider enforcing this requirement through an HTML linter or template helper. ]]>
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
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (28)

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 38)May include surrounding context.

md
<script src="api.html"></script>

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 61)May include surrounding context.

md
<script src="api.html"></script>

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 110)May include surrounding context.

md
<script src="api.html"></script>

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 47)May include surrounding context.

md
| **体验框架** | 复制 `experience-framework.html`,替换四维模型内容 |

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 60)May include surrounding context.

md
| **体验框架** | 复制 `experience-framework.html`,替换四维模型内容 |

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 194)May include surrounding context.

md
| **体验框架** | 复制 `experience-framework.html`,替换四维模型内容 |

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 · SKILL.md (reported line 71)May include surrounding context.

快速使用

html
<!-- Step 1: 引入配置 -->
<script src="api-console-config.js"></script>

<!-- Step 2: 修改配置 -->

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

Including an endpoint to obtain temporary COS credentials is sensitive because API consoles encourage direct invocation, and credential-issuing flows can grant access to storage resources if exposed to unintended users. In the context of a document-focused skill, this capability is especially risky because it is not obviously necessary and may normalize requesting secrets from a generic UI.

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 · assets/api.html (reported line 772)May include surrounding context.

html
</style>
</head>
<body class="creator-layout">
    <!-- 侧边栏 -->
    <aside class="icon-sidebar">
        <a href="create.html" class="icon-sidebar-logo" id="apiLogoLink"><span class="logo-wordmark" id="apiLogoText">本产品</span></a>
        <nav class="icon-nav">

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

Sensitive API keys are exposed in the DOM and can be copied with a single click, normalizing unsafe credential handling and enabling straightforward credential theft. Even if these are demo keys, publishing full key formats in a reusable skill encourages accidental deployment of real secrets in static assets.

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 · assets/api.html (reported line 1093)May include surrounding context.

html
</div>
    </main>

    <!-- 创建密钥弹窗 -->
    <div class="api-modal-overlay" id="createKeyModal">
        <div class="api-modal">
            <div class="api-modal-title">创建 API 密钥</div>

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 · assets/experience-framework.html (reported line 1253)May include surrounding context.

html
</head>
<body>

<!-- Scroll Progress -->
<div class="scroll-progress" id="scrollProgress"></div>

<!-- Left TOC -->

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 · assets/experience-framework.html (reported line 1949)May include surrounding context.

html
</tbody>
    </table>

    <!-- Test flow -->
    <h3 class="sub-heading reveal"><span class="sh-dot" style="background:var(--c-teal)"></span>测试流程</h3>

    <table class="data-table theme-teal reveal">

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill prominently advertises that the API Console can send real requests, but it does not present a clear, explicit safety warning about using production endpoints or triggering state-changing operations. In a documentation/debugging skill, that omission can lead users to unknowingly issue live requests against real services, causing accidental data creation, modification, or exposure.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The quick-start provides copy-paste instructions that immediately configure a base URL and render a request-sending console, but it lacks guardrails or warnings before users connect it to live infrastructure. Because the tool is intended for easy setup and real request execution, the missing warning increases the chance of unintended calls to production APIs.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The file ships a concrete, live API catalog for video processing, cloud storage operations, AIGC generation, and telemetry inside a skill described primarily as a document hub/template system. This materially expands the skill’s operational capability beyond passive documentation into real request-driving functionality, which increases the chance of unintended access, misuse, or surprise invocation against backend services.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

A temporary credential retrieval endpoint is presented like a normal API action, with no warning that the response may contain sensitive access material. Users of the console may invoke it casually, copy credentials into logs or screenshots, or expose them to unintended parties.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The telemetry endpoint invites transmission of event arrays without any user-facing notice about what data may be sent, whether it may include personal or behavioral data, or where it is stored. This creates privacy and compliance risk because users may unknowingly submit sensitive analytics payloads through the console.

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 declares lang="zh-CN" and all visible interface text is presented in Chinese, with no indication of user opt-in or a selectable locale. This can violate language/locale policy when a skill or interface forces a specific language without offering user choice or documenting a justified regional constraint.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
99% confidence
Finding

The page embeds full API keys directly in client-side HTML attributes, making them accessible to any visitor, browser extension, cached copy, or source inspection. In a documentation/API console skill, this is especially dangerous because users may assume the page is a safe template while it actually distributes live-looking credentials.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The create-key modal allows users to generate new API credentials, including full read/write scope, but it does not disclose that the resulting key is security-sensitive, may grant broad access, or should be copied and stored securely. Since this code implements credential creation without any confirmation text or cautionary notice, users are not warned about the risk of mishandling the secret.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
84% confidence
Finding

The manifest emphasizes unified documentation templates and a configurable API Console for sending requests, but this file also implements API credential lifecycle management UI, including creating and deleting keys. Credential administration is a materially broader capability than document templating or request debugging and is not clearly declared in the manifest description.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The flow describes direct upload of user videos to Tencent Cloud COS, processing via MPS, and later download/storage of rendered outputs, but it does not warn users that their media leaves the local device and is stored and processed by third-party cloud services. Because uploaded videos may contain sensitive or proprietary content, lack of notice can lead to unanticipated exposure, regulatory issues, and unsafe handling assumptions by downstream implementers.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The document explicitly states there is a '全自动埋点' system that automatically records the full user operation path, but it provides no user-facing notice, consent flow, retention details, or link to a privacy policy. In a product-document hub, describing comprehensive telemetry without transparency increases privacy and compliance risk because implementers may reproduce the design as-is and collect behavioral data without informing users.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
77% confidence
Finding

The inline comment indicates the key creation is only simulated, but the surrounding UI labels, buttons, and success messaging present it as actual API key creation and management. This creates an intent mismatch between developer-facing documentation and observable behavior of the page, especially on a screen labeled as API management.

Content

No source excerpt is available for this finding.