canvasflow-专业杂志级PPT一键生成专家

• Outcome: Review
• Latest audit: May 29, 2026 at 9:35 PM UTC
• Version:
  2.0.1

Overview

This presentation generator mostly matches its purpose, but its helper scripts automatically run macOS shell commands to open generated files, which creates a local execution risk.

Review before installing if you plan to run the bundled scripts, especially on macOS. Prefer removing the auto-open lines or changing them to a non-shell subprocess call behind an explicit --open option, and use trusted output paths. For sensitive presentations, consider pinning or removing third-party CDN assets before sharing or opening the generated HTML.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Behavioral AST: exec() Call, eval() Call, Dynamic Import
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration

Findings (13)

os.system() or os exec-family call

High

• Category: Dangerous Code Execution

Content

# 自动打开（macOS）
    if sys.platform == "darwin":
        os.system(f'open "{output_file}"')


if __name__ == "__main__":

• Confidence: 91% confidence
• Finding: os.system(f'open "{output_file}"')

os.system() or os exec-family call

High

• Category: Dangerous Code Execution

Content

# 自动打开（macOS）
    if sys.platform == "darwin":
        if results["html"]:
            os.system(f'open "{results["html"]}"')
        elif results["pptx"]:
            os.system(f'open "{results["pptx"]}"')

• Confidence: 90% confidence
• Finding: os.system(f'open "{results["html"]}"')

os.system() or os exec-family call

High

• Category: Dangerous Code Execution

Content

if results["html"]:
            os.system(f'open "{results["html"]}"')
        elif results["pptx"]:
            os.system(f'open "{results["pptx"]}"')


if __name__ == "__main__":

• Confidence: 90% confidence
• Finding: os.system(f'open "{results["pptx"]}"')

os.system() or os exec-family call

High

• Category: Dangerous Code Execution

Content

# 自动打开（macOS）
    if sys.platform == "darwin":
        os.system(f'open "{output_file}"')


if __name__ == "__main__":

• Confidence: 96% confidence
• Finding: os.system(f'open "{output_file}"')

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 85% confidence
• Finding: The skill instructs the agent to read local files, write generated content to disk, and execute a shell command (`open ...`) without declaring those capabilities as permissions. Undeclared powerful operations reduce transparency and can lead to unexpected file-system access or command execution in environments that rely on permission manifests for containment.

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 82% confidence
• Finding: The documentation overstates or misstates behavior, including hidden side effects such as writing files and invoking a shell command, while claiming capabilities that are not actually implemented. This mismatch can cause users or hosting platforms to grant trust under false assumptions, making risky operations more likely to occur without informed consent.

Context-Inappropriate Capability

Medium

• Confidence: 84% confidence
• Finding: Automatically launching the generated HTML is outside the core requirement of generating a presentation file and introduces an unnecessary execution capability. In a skill that writes user-influenced HTML, auto-opening the output can immediately trigger active content in the browser, making any embedded script or malicious markup more dangerous.

Context-Inappropriate Capability

Medium

• Confidence: 92% confidence
• Finding: The script automatically opens generated files on macOS without user confirmation, which exceeds the core generation function and can launch local applications unexpectedly. In a skill context, this is more dangerous because users expect content generation, not autonomous execution side effects that may open active content or external handlers.

Context-Inappropriate Capability

Low

• Confidence: 87% confidence
• Finding: The script automatically opens the generated file using a platform command even though file generation does not require code execution or launching external applications. This expands the script's side effects and can unexpectedly invoke local handlers, increasing user risk and making abuse of the unsafe shell invocation more consequential.

Vague Triggers

Medium

• Confidence: 92% confidence
• Finding: The trigger phrases include very generic requests such as “做 PPT” and “生成演示”, which are common in normal user conversations and can cause the skill to activate when the user did not explicitly intend to use this specific skill. In an agent environment, overly broad activation increases the chance of incorrect routing, unintended file generation, or the skill taking over unrelated presentation tasks without clear user consent.

Missing User Warnings

Medium

• Confidence: 88% confidence
• Finding: The skill executes a shell command to open the generated HTML automatically, without confirmation. This removes user control and can cause unintended execution of browser-side scripts or external resource loads as soon as the file is created, which is riskier in this context because multiple slide fields are inserted into HTML without escaping.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: Automatic shell-based execution without prior warning violates the principle of least surprise and creates an avoidable execution pathway. Even if intended for convenience, using shell execution to launch generated artifacts can be abused or can expose users to unintended application behavior, especially when file content may be partially derived from user-controlled input.

Missing User Warnings

Medium

• Confidence: 83% confidence
• Finding: Automatically executing a shell command without prior confirmation is unsafe because it performs an external action the user did not explicitly approve. In this skill's context, the risk is amplified by the fact that the opened target path is user-influenced and the action is unrelated to the core requirement of generating a presentation file.

VirusTotal

66/66 vendors flagged this skill as clean.