Back to skill

Security audit

PRD to Prototype

Security checks for vulnerabilities and agentic risk

Overview

This skill is a coherent PRD/prototype generator, but it can trigger from broad product-idea phrasing, write files immediately, and deploy externally without a clearly separate user consent step.

Install only if you are comfortable with the skill activating from general product-idea requests, creating files under /workspace, and potentially deploying a generated prototype. Review generated PRDs before sharing, and avoid using sensitive or confidential product concepts unless deployment is disabled or explicitly confirmed.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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 (1)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:161
Finding

Unpinned Remote Dependencies Without Subresource Integrity

Content
View full analysis
``` ### Technical Analysis The mandatory prototype template loads Tailwind CSS as executable JavaScript from the mutable `https://cdn.tailwindcss.com` endpoint. It also loads Font Awesome from a third-party CDN. Neither resource is protected with a Subresource Integrity hash, and the Tailwind URL does not identify an immutable dependency version. Consequently, the content executed by generated prototypes may change after the Skill has been reviewed. If the CDN, its publishing account, DNS resolution, or the delivery chain is compromised, an attacker could cause arbitrary JavaScript to execute when a user opens a generated prototype. The external image usage documented elsewhere in the Skill is expected presentation behavior and was not identified as a vulnerability. The security concern here is specifically the unverified remote dependency, especially the executable Tailwind script. ### Attack Path 1. A user invokes the Skill and confirms prototype generation. 2. The Agent creates an HTML prototype using the mandatory template from `SKILL.md`. 3. The generated prototype is opened locally or deployed to a web origin. 4. The browser requests the Tailwind script from the external CDN. 5. An attacker who has compromised the CDN, upstream publishing process, or relevant delivery infrastructure serves modified JavaScript. 6. The attacker-controlled script executes in the prototype's browser context. 7. The script may inspect or modify the DOM, capture information entered into the prototype, access origin-readable browser storage, and issue requests with ...[truncated 822 chars]
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (5)

Vague Triggers

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The manifest description advertises generic trigger phrases like '我想做一个...' and '帮我设计...', which overlap with ordinary conversation and many unrelated design/product requests. In an agent environment, such vague activation language can cause accidental routing into a workflow that writes artifacts and advances execution without sufficiently clear user intent.

Content

No source excerpt is available for this finding.

Vague Triggers

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The trigger condition is extremely broad: any user expressing a product idea, need, pain point, or creativity can activate the skill. That creates a high risk of unintended invocation, causing the agent to begin generating files and workflow outputs in contexts where the user did not explicitly request this skill or consent to side effects.

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 · SKILL.md (reported line 160)May include surrounding context.

md
<meta charset="UTF-8">
    <meta name="viewport" content="width=device-width, initial-scale=1.0">
    <title>产品名称</title>
    <!-- 引入 Tailwind -->
    <script src="https://cdn.tailwindcss.com"></script>
    <!-- 引入 FontAwesome -->
    <link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/6.4.0/css/all.min.css">

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill instructs writing /workspace/docs/prd.md immediately, without any user-facing notice that files will be created or modified. Because the skill also enforces 'zero-question' execution, users may not realize system state is being changed merely by discussing an idea.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The workflow calls for deployment whenever a deploy tool is available, but provides no requirement for informed user consent about publication or external exposure. This could unintentionally publish sensitive product concepts, internal workflows, or branded mockups to an external endpoint.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.