Back to skill

Security audit

Tool Showcase PPT By HTML

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a coherent HTML presentation generator, but generated decks automatically load mutable third-party JavaScript and external fonts despite being presented as single-file HTML.

Review generated decks before sharing or opening in sensitive contexts. Prefer removing the CDN script, vendoring a pinned Lucide release, using local/system fonts, or adding exact version and integrity controls so the presentation does not automatically trust mutable third-party code.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Findings (1)

T03 · Remote Payload Retrieval and Execution

Warning
Location
assets/template.html:911
Finding

Unpinned Third-Party JavaScript Is Retrieved and Executed at Runtime

Content
View full analysis

Vulnerability Details

File Location: assets/template.html:911
Vulnerability Type: Mutable remote JavaScript dependency without version pinning or Subresource Integrity
Risk Level: Medium

Vulnerable Code

html
<script src="https://unpkg.com/lucide@latest/dist/umd/lucide.min.js"></script>

Technical Analysis

The presentation template automatically retrieves and executes the Lucide UMD bundle from the unpkg CDN whenever a generated HTML presentation is opened with network access.

The dependency uses the mutable @latest tag rather than an exact package version. It also lacks a Subresource Integrity hash. Consequently, the JavaScript executed by generated presentations can change after the Skill package has been reviewed. The effective runtime behavior therefore depends on the security of the upstream package, its maintainer account, package registry infrastructure, and CDN.

This is a supply-chain trust-boundary weakness. If any relevant upstream component is compromised or a malicious release becomes the version resolved by @latest, arbitrary attacker-controlled JavaScript could execute in the presentation's browser context.

No evidence indicates that the currently referenced Lucide package is malicious. The vulnerability is the unsafe and mutable method by which executable third-party content is loaded.

Attack Path

  1. An attacker compromises the upstream package, a maintainer account, the package publication process, or the CDN delivery path.
  2. The attacker causes lucide@latest to resolve to a modified JavaScript bundle.
  3. A user opens an HTML presentation generated from this template while connected to the network.
  4. The browser requests the script from unpkg.
  5. Because there is no exact version or integrity check, the browser accepts and executes the modified response.
  6. The malicious script runs in the generated presentation's browser context.

Impact Assessment

Successful exploitation permits arbitrary ...[truncated 795 chars]

Remediation
View remediation

Remediation Suggestions

  1. Preferred: bundle the dependency locally
    • Download and review a specific Lucide release.
    • Store the reviewed JavaScript inside the project.
    • Reference it using a relative path, for example:
html
<script src="assets/vendor/lucide-0.x.y.min.js"></script>
  1. If remote loading is unavoidable, pin an exact version

    • Replace @latest with a fixed, reviewed version.
    • Do not use mutable tags or unversioned URLs.
  2. Add Subresource Integrity

    • Generate and verify a cryptographic hash for the exact dependency.
    • Add both integrity and crossorigin attributes:
html
<script
  src="https://unpkg.com/lucide@EXACT_VERSION/dist/umd/lucide.min.js"
  integrity="sha384-REVIEWED_HASH"
  crossorigin="anonymous"></script>
  1. Apply a restrictive Content Security Policy

    • Prefer a policy that permits scripts only from the generated document itself.
    • If a CDN must remain enabled, allow only the required host and avoid permissive directives such as unsafe-eval.
  2. Document and monitor dependency provenance

    • Record the selected version and hash.
    • Review dependency updates before changing the pinned artifact.
    • Consider automated integrity verification as part of release checks.
  3. Consider removing the runtime dependency

    • Pre-render the required SVG icons into generated presentations, eliminating third-party runtime JavaScript entirely.
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (13)

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 217)May include surrounding context.

md
`template.html` 是**完整可运行**的文件,CSS、翻页 JS、Lucide 图标 CDN 全已预设,只有 `<!-- SLIDES_HERE -->` 占位符等待填充。

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 362)May include surrounding context.

md
`template.html` 是**完整可运行**的文件,CSS、翻页 JS、Lucide 图标 CDN 全已预设,只有 `<!-- SLIDES_HERE -->` 占位符等待填充。

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 · references/layouts.md (reported line 1560)May include surrounding context.

在页面中需要数字滚动的地方,使用 .num-roll 类 + data-target 属性:

html
<!-- 在 .stat-value 或 .metric-value 中使用 -->
<span class="stat-value num-roll" data-target="200">0</span>
<span class="unit">ms</span>

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 · references/layouts.md (reported line 1560)May include surrounding context.

在页面中需要数字滚动的地方,使用 .num-roll 类 + data-target 属性:

html
<!-- 在 .stat-value 或 .metric-value 中使用 -->
<span class="stat-value num-roll" data-target="200">0</span>
<span class="unit">ms</span>

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The HTML root sets lang="zh-CN", and the visible template text is also Chinese, which makes the template default to a specific language/locale. Under the policy, forcing a locale without user opt-in or a documented justification is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The template loads fonts from Google Fonts, so the generated PPT is not actually self-contained and will make outbound network requests when opened. This creates privacy, availability, and supply-chain risk: viewing the file leaks metadata to third parties and the presentation can break or change if the external resource is unavailable or modified.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The template imports JavaScript from unpkg at runtime, which gives an external CDN influence over code executed when the HTML file is opened. In a skill advertised as a single-file PPT, this is especially risky because it introduces a remote script supply-chain dependency, breaks offline use, and can expose viewers to malicious script changes or tracking if the CDN or package path is compromised.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The text states that the tool-introduction PPT should avoid large blocks of Chinese body text and caps titles at a number of Chinese characters, which imposes a specific language/locale assumption. Because the file does not offer user opt-in or explain that this skill is intended only for a Chinese-language context, this is a natural-language locale policy concern.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

This markdown file is written throughout in Chinese and presents the templates and instructions as the default interaction/output language, without any indication that users may choose another language or locale. Under the policy rule, forcing a specific language without user opt-in is a natural-language policy concern.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The file is written entirely in Chinese and repeatedly instructs the AI how to behave, but nowhere offers an opt-in or alternative language/locale choice. Under the policy, forcing a specific language without user choice is a natural-language policy violation unless the locale constraint is explicitly documented and justified.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

The instruction "Inter + Noto Sans SC,任何衬线都是错的" imposes a specific language/script-oriented typography choice, including Simplified Chinese font usage, as an absolute requirement. This is a natural-language policy concern because it removes user choice on locale-appropriate presentation and is not documented as a region-specific constraint.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

The checklist recommends that h-hero titles be limited to 8 Chinese characters, which is a locale-specific instruction. Without stating that the skill is exclusively for Chinese presentations or offering alternatives for other languages, this creates a language policy issue.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

The self-check includes a hard requirement that main titles be no more than 8 Chinese characters, reinforcing a mandatory Chinese-language assumption. This is a locale-specific policy constraint that is not framed as optional or limited to a Chinese-only workflow.

Content

No source excerpt is available for this finding.