Subtitle Generator

• Outcome: Review
• Latest audit: May 29, 2026, 9:29 PM
• Version:
  1.0.7

Overview

This subtitle skill appears intended to generate subtitles, but it automatically modifies the local environment and gives broad background/process instructions that deserve user review before installation.

Install only if you are comfortable with a background subtitle tool that may create ~/.whisper-venv, download Python packages, write subtitle files beside your videos, and send completion notifications containing local paths. Prefer running it in a controlled working directory, and look for a revised version with pinned dependencies, explicit setup consent, unique temporary files, and PID-scoped cleanup.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• Behavioral AST: exec() Call, eval() Call, Dynamic Import
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands

Findings (9)

subprocess module call

Medium

• Category: Dangerous Code Execution

Content

python_cmd = shutil.which("python") or shutil.which("python3") or "python3"
        else:
            python_cmd = shutil.which("python3") or shutil.which("python") or "python3"
        result = subprocess.run(
            [python_cmd, "-m", "venv", str(venv_path)],
            capture_output=True,
            text=True,

• Confidence: 79% confidence
• Finding: result = subprocess.run( [python_cmd, "-m", "venv", str(venv_path)], capture_output=True, text=True, timeout=120, )

subprocess module call

Medium

• Category: Dangerous Code Execution

Content

# Step 3: Install requirements.txt
    print("[Setup] Installing dependencies...")
    try:
        result = subprocess.run(
            [str(pip_path), "install", "-r", str(requirements_path)],
            capture_output=True,
            text=True,

• Confidence: 91% confidence
• Finding: result = subprocess.run( [str(pip_path), "install", "-r", str(requirements_path)], capture_output=True, text=True, timeout=600, )

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 94% confidence
• Finding: The skill clearly invokes shell commands and writes files, but it declares no permissions. This creates a transparency and policy-enforcement gap: a user or platform may approve the skill expecting a harmless subtitle utility while it can execute commands, create environments, install packages, and modify files on disk.

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 97% confidence
• Finding: The documented purpose is subtitle generation, but the skill also performs environment bootstrapping, dependency installation, and system-event based wake/notification behavior. That mismatch is dangerous because users may trigger it for transcription without realizing it will execute package-management and environment-configuration actions on the host.

Context-Inappropriate Capability

Medium

• Confidence: 94% confidence
• Finding: The skill's manifest describes subtitle/ASR functionality, but the implementation silently bootstraps a virtual environment and executes Python/pip subprocesses. That mismatch increases risk because users would not reasonably expect environment provisioning and package installation as part of routine subtitle generation.

Context-Inappropriate Capability

Medium

• Confidence: 88% confidence
• Finding: The skill includes an external system-event notification feature that is not necessary for subtitle generation and can leak file names, paths, and processing metadata to another local tool. In an agent setting, unrelated side effects are more dangerous because they expand the trust boundary and create an unexpected exfiltration or signaling channel.

Intent-Code Divergence

Medium

• Confidence: 83% confidence
• Finding: The help text claims the main session AI will automatically send a Telegram notification, but the code only emits an openclaw system event. In an agent environment, misleading capability descriptions are dangerous because they obscure where data is sent and may trick operators into approving behavior they do not fully understand.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: The skill omits a clear warning that first use may create a virtual environment and auto-install dependencies. This is dangerous because users can unknowingly authorize network access, package installation, disk writes, and environment changes simply by asking for subtitles.

Missing User Warnings

Medium

• Confidence: 97% confidence
• Finding: The code creates `~/.whisper-venv` and installs dependencies automatically without any prior warning or confirmation. Even if intended for convenience, silent filesystem and environment modification is a security and trust issue, especially for a user-facing subtitle skill where such side effects are not clearly disclosed.

VirusTotal

65/65 vendors flagged this skill as clean.