Back to skill

Security audit

Whiteboard Animation

Security checks for vulnerabilities and agentic risk

Overview

The skill appears to do what it claims, but it will create a local Python environment and install third-party media packages when needed.

Install this only if you are comfortable with it creating a local .venv and downloading unpinned Python packages. Prefer running it in an isolated environment, review or pin dependencies first if supply-chain control matters, and confirm explicit image paths/output locations before running batch jobs.

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
scripts/setup_env.py:22
Finding

Automatic Installation of Unpinned Third-Party Dependencies

Content
View full analysis

Vulnerability Details

File Location: scripts/setup_env.py:22-27 and scripts/setup_env.py:60-69; the installation workflow is invoked by SKILL.md:29-40
Vulnerability Type: Supply-chain risk caused by unpinned automatic dependency installation
Risk Level: Medium

Vulnerable Code

python
REQUIRED_PACKAGES = {
    "cv2": "opencv-python",
    "numpy": "numpy",
    "av": "av",
}
python
def install_packages(python_path, packages):
    """Install the specified packages in the virtual environment."""
    if not packages:
        return True
    print(f"[..] Installing dependencies: {', '.join(packages)}")
    result = subprocess.run(
        [python_path, "-m", "pip", "install", "--quiet"] + packages,
        capture_output=True,
        text=True,
    )
    if result.returncode != 0:
        print(f"[Error] Installation failed:\n{result.stderr}")
        return False
    print("[OK] Dependency installation completed")
    return True

The original source contains Chinese display strings, but the executable behavior represented above is unchanged: package names are passed directly to pip install without version or integrity constraints.

Technical Analysis

The documented workflow instructs the agent to run setup_env.py when the required environment is unavailable. The script installs opencv-python, numpy, and av using their bare package names. It does not specify exact versions, cryptographic hashes, an approved package index, or a reviewed lock file.

As a result, the effective code installed and imported by the Skill can change after the Skill itself has been audited. Pip also uses the environment's configured indexes and mirrors, which may differ between systems. A compromised package release, transitive dependency, package index, or locally configured mirror could therefore introduce attacker-controlled code.

Installation and subsequent imports ex ...[truncated 1718 chars]

Remediation
View remediation

Remediation Suggestions

  1. Create a reviewed dependency manifest containing exact versions for every direct and transitive dependency.

  2. Generate and verify cryptographic hashes for all approved distributions.

  3. Install dependencies using a command such as:

    bash
    python -m pip install --require-hashes -r requirements.txt
    
  4. Configure an explicit trusted package index or an internally controlled artifact repository rather than relying on ambient pip configuration.

  5. Review and update pinned dependencies through a controlled process that includes vulnerability scanning and compatibility testing.

  6. Prefer provisioning dependencies during an explicit installation phase rather than automatically downloading code during normal Skill execution.

  7. If automatic installation must remain, fail closed when the lock file, hashes, approved index, or expected package versions cannot be verified.

  8. Run dependency installation and media processing in a least-privileged, network-restricted environment with access limited to necessary input and output directories.

Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (9)

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
96% confidence
Finding

The skill instructs the agent to run shell commands, create a virtual environment, install dependencies, and read/write local files, but it declares no explicit tool scope or permissions. This increases the blast radius because an agent may invoke filesystem and shell capabilities without clear least-privilege boundaries or user-visible consent expectations.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The trigger phrase “白板动画” is overly broad and can match casual conversation, requests for explanation, or unrelated discussion rather than an explicit intent to execute the skill. That can cause unintended activation of a workflow that performs shell execution, dependency installation, and file operations, making accidental tool use more likely.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

整个技能文件以中文规定名称、描述和操作说明,但未说明这是区域性/合规性要求,也未给出用户可选择语言的机制。按规则,若技能强制特定语言而没有用户选择或明确正当理由,属于自然语言政策问题。

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The module docstring and command-line description/help text are written entirely in Chinese, which imposes a specific language on users. There is no indication that Chinese is optional, configurable, or required for a documented region-specific purpose.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/batch_generate.py (reported line 86)May include surrounding context.

python
if args.no_hand:
            cmd.append("--no-hand")

        result = subprocess.run(cmd)
        success = result.returncode == 0
        results.append({
            "image": image,

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The module docstring and all user-facing CLI/help/status text are written in Chinese, which effectively imposes a specific language on users. The file does not provide an opt-in language selection or explain that the tool is intended only for a Chinese-speaking or region-specific context.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/setup_env.py (reported line 52)May include surrounding context.

python
def check_package(python_path, import_name):
    """检测某个包是否可在虚拟环境中导入"""
    result = subprocess.run(
        [python_path, "-c", f"import {import_name}"],
        capture_output=True,
    )

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/setup_env.py (reported line 64)May include surrounding context.

python
if not packages:
        return True
    print(f"[..] 安装依赖: {', '.join(packages)}")
    result = subprocess.run(
        [python_path, "-m", "pip", "install", "--quiet"] + packages,
        capture_output=True,
        text=True,

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
94% confidence
Finding

This code file contains natural-language documentation and runtime messages entirely in Chinese, which imposes a specific language on users without any opt-in or fallback. The policy for this category flags language or locale constraints when the skill does not offer a language choice or clearly justify the restriction.

Content

No source excerpt is available for this finding.