Back to skill

Security audit

prd-design-iteration

Security checks for vulnerabilities and agentic risk

Overview

This is a coherent product-design workflow that saves user-provided product details and screenshots locally to produce a PRD and prototype, with privacy and dependency cautions but no evidence of deception or exfiltration.

Install only if you are comfortable with a Chinese-language workflow that writes product notes, screenshots, and generated prototypes into the local project workspace. Redact customer data, credentials, internal URLs, and production secrets before uploading screenshots or HTML. Open generated prototypes in an isolated browser context where possible because the template loads third-party CDN scripts.

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
references/templates/prototype-template.html:11
Finding

Unpinned Third-Party Scripts Without Subresource Integrity

Content
View full analysis

Vulnerability Details

File Location: references/templates/prototype-template.html, lines 11–14
Vulnerability Type: Third-party dependency supply-chain risk
Risk Level: Medium

Vulnerable Code

html
<script crossorigin src="https://unpkg.com/react@18/umd/react.production.min.js"></script>
<script crossorigin src="https://unpkg.com/react-dom@18/umd/react-dom.production.min.js"></script>
<script src="https://unpkg.com/@babel/standalone/babel.min.js"></script>
<script src="https://cdn.tailwindcss.com"></script>

Technical Analysis

The executable prototype template imports four JavaScript resources from external CDNs. React and ReactDOM use mutable major-version selectors, while Babel and Tailwind do not specify versions. None of these resources has a Subresource Integrity hash.

When a generated prototype is opened, the browser retrieves and executes the current content returned by these external services. Consequently, the effective executable code can change after the Skill package has been reviewed. A compromised CDN, upstream package, maintainer account, DNS route, or mutable release selector could cause attacker-controlled JavaScript to execute in the prototype page.

This is limited to prototypes generated from or based on this template. The audit found no evidence that the current external dependencies are intentionally malicious.

Attack Path

  1. An attacker compromises an upstream package account, CDN asset, delivery infrastructure, or other component in the dependency supply chain.
  2. The attacker modifies a referenced script or causes a mutable version selector to resolve to a malicious release.
  3. A user generates or opens a prototype based on this template while connected to the network.
  4. The browser downloads the modified JavaScript because the template does not enforce an expected integrity hash.
  5. The malicious script executes in the prototype's browser context.
  6. It can modify the displayed prototype, read i ...[truncated 907 chars]
Remediation
View remediation

Remediation Suggestions

  1. Pin every dependency to an exact, reviewed version rather than using a major-version selector or an unversioned URL.
  2. Add cryptographic Subresource Integrity hashes and appropriate crossorigin="anonymous" attributes to externally hosted static resources.
  3. Prefer vendoring reviewed dependencies locally so generated prototypes do not execute mutable remote code.
  4. Replace runtime Babel transformation and Tailwind CDN compilation with a controlled build process that produces static JavaScript and CSS artifacts.
  5. Regularly review and update pinned dependencies through a documented dependency-management process.
  6. Host generated prototypes on an isolated origin that does not share cookies, storage, or privileges with sensitive applications.
  7. Apply a restrictive Content Security Policy that permits scripts only from approved sources and blocks unexpected outbound connections.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (10)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The trigger description includes broad phrases like “改一下” and “or similar iteration requests,” which can match many unrelated user requests and cause unintended invocation of this skill. In an agent environment, overbroad routing can lead to wrong workflow selection, unnecessary file creation, and collection of screenshots or internal process details the user did not intend to provide to this skill.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
84% confidence
Finding

The skill content forces a Chinese-language interaction context without asking for user preference, which can degrade reliability and cause misunderstanding if the user's requested language differs. While this is not a direct code-execution issue, it can produce incorrect outputs, failed confirmations, or accidental disclosure of sensitive internal workflow details due to miscommunication.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

All user-facing prompts and instructions in this file are fixed in Chinese, and there is no indication that the user can choose another language or that the skill is intentionally restricted to a Chinese-speaking context. This can violate language/locale policy when a specific language is imposed without user opt-in.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The confirmation criteria include very broad phrases like “没问题” and “对的”, which commonly appear in conversational feedback that may not indicate informed approval to proceed. In this skill, that can cause the agent to advance to Step 3 without a reliable, explicit user confirmation, leading to unintended product-spec or prototype changes based on incorrect module inference.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill instructs the agent to save user-uploaded screenshots to disk, but it provides no warning that screenshots may contain sensitive internal information such as customer data, credentials, environment details, or proprietary workflows. In a company-internal product design context, retaining these files locally without explicit consent, minimization, or handling guidance creates a real privacy and data exposure risk.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill explicitly requests current UI screenshots or HTML prototypes from an internal system and directs saving them locally, again without any notice about retention, sensitivity, or safe handling. HTML prototypes can also embed internal endpoints, tokens, comments, or business logic details, so collecting and persisting them without safeguards increases the chance of unintended disclosure.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The document title and all procedural instructions are written in Chinese and direct the agent to produce outputs such as validation reports and PRD artifacts in that language. There is no indication that the user may choose another language or that the locale restriction is required for a region-specific purpose, which creates a language policy concern.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

标题及全文内容均以中文固定呈现,未见任何说明允许用户选择语言或表明该清单仅适用于特定中文场景。根据规则,若技能在自然语言层面强制特定语言且无用户选择或正当限定,可构成语言/locale 策略问题。

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
94% confidence
Finding

The step explicitly instructs creating project/00-input/baseline.md and presenting it, but provides no user disclosure or consent boundary around writing user-supplied business information to disk. In an internal process-management context, that file may contain sensitive operational details, so silent persistence increases privacy, retention, and data-handling risk even if the content is not directly code-executable.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
95% confidence
Finding

The HTML document sets lang="zh-CN", and the visible UI text throughout the template is written only in Chinese. This imposes a specific language/locale by default without any user opt-in or explanation that the template is intended exclusively for a China-specific context.

Content

No source excerpt is available for this finding.