Security audit

subtitle-extractor

Security checks for vulnerabilities and agentic risk

Overview

This skill performs the subtitle extraction and transcription workflow it advertises, but users should understand that it may use cookies, download media/model files, and leave temporary transcript or audio files on disk.

Install only if you are comfortable running yt-dlp/faster-whisper locally, allowing video and model downloads, and providing platform cookies when needed. Use narrowly scoped cookie exports, avoid private local media unless necessary, and clear temporary audio/subtitle files after sensitive jobs.

SkillSpector

Vulnerability Patterns

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
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (5)

Lp3

Medium

Category: MCP Least Privilege
Confidence: 96% confidence
Finding: The skill instructs the agent to run shell commands, read environment variables, discover local cookie files, and write output files, but its metadata declares only binary requirements and omits explicit permissions for shell, file, and environment access. This creates a trust and policy gap: an execution framework may permit behavior users did not clearly consent to, including access to sensitive local files such as cookie exports and creation of files on disk.

Missing User Warnings

Low

Confidence: 84% confidence
Finding: The skill makes local file creation mandatory by directing the agent to save subtitle content into the outputs directory, but it does not clearly warn the user beforehand that files will be created and renamed on disk. While expected for this workflow, undisclosed file writes can surprise users, overwrite assumptions about workspace state, or expose subtitle content in shared environments.

Missing User Warnings

Medium

Confidence: 87% confidence
Finding: The script writes generated subtitles to a predictable persistent temp location and does not remove them afterward or warn the user that potentially sensitive transcript content will remain on disk. Subtitle files may contain private or copyrighted material, and temp directories are often accessible to other local users, backup tools, or later processes.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: Downloaded audio is stored in a persistent temp file outside the TemporaryDirectory lifecycle, and the code does not clean it up or notify the user. Media content can be highly sensitive, and leaving it in a shared temp area increases exposure to local disclosure, unintended retention, and forensic recovery.

Missing User Warnings

Medium

Confidence: 87% confidence
Finding: The default workflow persists extracted or transcribed subtitles to a temp file without warning, which can expose sensitive speech content or personal information contained in the transcript. Because this is the primary execution path, users may unknowingly leave recoverable transcript artifacts on disk.

Static analysis

No suspicious patterns detected.