Back to skill

Security audit

会议纪要融合生成器

Security checks for vulnerabilities and agentic risk

Overview

This Chinese meeting-minutes skill is coherent and purpose-aligned, with normal privacy caution for meeting files and generated outputs.

Use this skill for Chinese-language meeting-minutes generation from PDFs you intentionally provide. Confirm the exact source files and destination folder before running it, and only export HTML that was generated from trusted or properly escaped meeting content.

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 (1)

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/export_pdf.py:17
Finding

Untrusted HTML Rendered with Active Browser Capabilities

Content
View full analysis
/assets/template.html` 为设计系统参考,生成新的 HTML 文件。 模板中的设计令牌已经封装在 CSS 变量中,跨平台兼容: ``` ### Technical Analysis The exporter accepts an arbitrary HTML path from the command line and opens that document in Playwright Chromium. The browser context retains its default JavaScript and networking capabilities. No request interception, origin restriction, Content Security Policy, markup sanitization, or JavaScript disabling is applied. Consequently, an input document containing active HTML can execute JavaScript and issue outbound requests while the PDF is being generated. This is directly exploitable when an attacker can provide or modify the HTML passed to the script. A secondary path may exist when attacker-controlled meeting content is copied into generated HTML without context-aware escaping, as the Skill instructions do not mandate escaping or sanitization. The packaged `assets/template.html` itself contains no scripts or external URLs; therefore, exploitation requires malicious or compromised input rather than occurring from the bundled template alone. ### Attack Path 1. An attacker supplies a crafted HTML document or causes malicious markup to be included in meeting-derived content. 2. The operator or agent invokes: ```bash python3 scr ...[truncated 1321 chars]
Remediation
View remediation
Vulnerability Patterns
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (7)

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

There is a clear description-behavior mismatch. The declared purpose describes an end-to-end meeting-minutes processing pipeline: ingesting source PDFs from specific platforms, extracting and reconciling content, correcting ASR mistakes, and producing a polished structured output. The provided code chunk does none of that substantive work. It only converts an already-existing HTML file into a PDF, preferring Playwright and falling back to WeasyPrint. While PDF export could be a supporting implementation detail of the described skill, this chunk by itself does not match the declared primary functionality and omits the core advertised capabilities.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

L003 将“会议纪要、meeting minutes、纪要模板、生成纪要、整理会议记录、会议总结”等大量泛化表述都作为触发条件,其中多项属于常见办公语境里的普通表达。该描述没有限定必须是处理钉钉/得到 PDF 纪要,也没有给出不应触发的反例,容易在普通总结或写作请求中误调用此技能。

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill instructs the agent to read meeting-note PDFs and write generated HTML/PDF files to the user's working directory without explicit privacy, consent, or destination-confirmation safeguards. Because meeting minutes often contain sensitive business content, silent processing or writing to an unintended directory can expose confidential data or create accidental data retention.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The skill directs execution of a local Python export script and browser automation to render PDFs, but does not require an explicit approval gate for code execution or explain the risks of running local scripts and headless browser tooling. In a sensitive environment, this can lead to unintended execution of unreviewed code, file access, or network-capable rendering behavior beyond what the user expected.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The HTML root sets lang="zh-CN", which establishes a fixed Chinese locale for the document. Because the file contains no accompanying notice that this template is intentionally region-specific or that users may choose another language, it appears to impose a specific language/locale without opt-in.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
78% confidence
Finding

文件整体将输出、命名、版式和示例均固定为中文语境,如中文标题、中文文件名格式及中文字体链,但未说明是否支持用户选择其他语言或 locale。对于通用会议纪要技能,这构成了默认强制特定语言/地区呈现的倾向。

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
93% confidence
Finding

This code file contains natural-language documentation in Chinese only, describing the script purpose and usage without offering any language or locale choice. Under the policy, forcing a specific language without user opt-in can be a natural-language policy violation when no justification or alternative is provided.

Content

No source excerpt is available for this finding.