Back to skill

Security audit

教育资源下载助手

Security checks for vulnerabilities and agentic risk

Overview

The skill is not clearly malicious, but it gives practical help for bulk course downloading, stream capture, login-cookie use, and unverified tool installs.

Install only if you are comfortable with a Review-level skill that may guide bulk downloading from education platforms. Use official download features first, do not share cookies or login tokens, avoid bypassing platform restrictions, and verify any third-party tool source, version, and dependencies before running local install commands.

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:136
Finding

Unverified and Unpinned Third-Party Dependencies

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 136 and 146–153; additional references at lines 228–230
Vulnerability Type: Supply-chain risk from unverified and unpinned dependencies
Risk Level: Medium

Relevant code snippets:

bash
pip install yt-dlp
bash
git clone https://github.com/xxx/knowledge-grab.git
cd knowledge-grab
pip install -r requirements.txt

git clone https://github.com/xxx/xuetangx-downloader.git
cd xuetangx-downloader
pip install -r requirements.txt

Additional tool references include:

text
yt-dlp: pip install yt-dlp
IDM: idmchina.net
JDownloader2: jdownloader.org

Technical Analysis

The skill instructs users to install an unpinned package from PyPI and execute dependency manifests obtained from external repositories. No package versions, cryptographic hashes, signatures, immutable commit identifiers, or verified publisher information are supplied.

The repository URLs use the placeholder owner xxx. Because these examples do not identify authoritative repositories, users may search for similarly named projects and select an impersonated, abandoned, or compromised repository. Running pip install -r requirements.txt then processes dependency declarations controlled by that repository. Python packages can execute code during installation or when the downloaded application is subsequently launched.

The external software references also lack documented integrity-verification procedures. This creates a dependency-confusion, typosquatting, repository-substitution, and upstream-compromise risk. The invalid placeholder URLs reduce direct exploitability of the displayed commands, but the surrounding instructions encourage users to discover and execute unverified alternatives.

Attack Path

  1. An attacker publishes a repository or package using one of the recommended or confusingly similar names.
  2. A user follows the skill, searches for ...[truncated 1199 chars]
Remediation
View remediation

Remediation Suggestions

  1. Replace all placeholder repository URLs with verified, authoritative upstream repositories.
  2. Document the expected publisher, repository owner, and package identity so users do not rely on search results alone.
  3. Pin Python packages and transitive dependencies to reviewed versions using a lock file.
  4. Require hashes for downloaded distributions, such as with pip install --require-hashes -r requirements.txt.
  5. Pin cloned repositories to reviewed commit hashes or signed release tags rather than mutable default branches.
  6. Provide expected SHA-256 checksums or signature-verification instructions for installers and binary downloads.
  7. Recommend installation inside a dedicated virtual environment or sandbox using a non-administrative account.
  8. Require review of requirements.txt, package metadata, installation hooks, and downloaded scripts before execution.
  9. Link only to official publisher pages and advise users not to install similarly named tools from unverified search results.
  10. Periodically review pinned dependencies for known vulnerabilities and update them through a controlled process.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
Findings (7)

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

Although framed as compliant use, the skill includes troubleshooting for 403 blocks, login-cookie refresh, encrypted m3u8 handling, User-Agent tuning, and rate-limit reduction. These are common anti-abuse evasion and session-maintenance techniques that materially assist bypassing platform restrictions and sustaining unauthorized automated access.

Content

No source excerpt is available for this finding.

Ssd 4

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The overall document uses a safety/compliance framing to legitimize a workflow that still teaches practical content extraction at scale, including bulk download tooling, m3u8 discovery, download scripting, anti-block troubleshooting, and login-state maintenance. This framing can reduce user caution while preserving the core harmful capability of systematic acquisition from protected platforms.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill claims it will not generate crawler code, but it still instructs the model to create automation scripts, configure download tools, and use traffic-capture workflows to acquire course content in bulk. That contradiction lowers safety assurances and enables operational assistance for large-scale extraction of protected materials even without providing a full crawler.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The skill instructs users on high-risk actions like configuring User-Agent strings, capturing video links, and updating login cookies without warning about credential theft, token leakage, account compromise, or privacy exposure. In practice, this can lead users to paste sensitive session material into chats or misuse authentication data in unsafe tooling.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

该技能是 markdown 文件,因此需要检查模糊触发条件。L247 将“下载教育课程”作为触发场景,表述过于泛化,可能匹配大量普通学习类请求,而不仅限于本技能面向的特定平台下载与配置场景。

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
75% confidence
Finding

规则要求检查自然语言层面的语言或地区策略问题。该技能整体以中文编写并默认中文交互,没有看到明确的语言选择或用户 opt-in 说明,可能构成默认强制特定语言。

Content

No source excerpt is available for this finding.

Vague Triggers

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

该文件给出了若干正向触发示例,但没有提供负面示例、适用范围限制或排除条件。根据规则,缺少触发范围约束或负例说明会增加误触发风险。

Content

No source excerpt is available for this finding.