Security audit

Video Analyzer

Security checks for vulnerabilities and agentic risk

Overview

This skill does what it advertises: it downloads user-supplied videos, transcribes them locally, extracts frames, and prepares transcript summaries, with no hidden persistence or deceptive behavior found.

Install only if you are comfortable with a skill that downloads remote videos, runs yt-dlp/ffmpeg/Whisper locally, creates media and transcript files, and may require substantial disk, CPU, and model-download bandwidth. Review transcripts before pasting or piping them to any AI service, especially for meetings, private recordings, credentials, or confidential business content.

A.I.G

Vulnerability Patterns

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
Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (6)

Tp4

High

Category: MCP Tool Poisoning
Confidence: 96% confidence
Finding: The declared description presents a media-processing skill focused on obtaining and analyzing video content through downloading, transcription, and frame extraction. The actual code chunk is a summarization helper: it validates a transcript file, reads its contents, and emits a structured prompt for manual or CLI-based AI summarization. This is a materially different behavior from the stated purpose in this specific code chunk, and the core declared capabilities are absent.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The script accepts an arbitrary user-supplied URL, downloads remote media with yt-dlp, and then parses the resulting content with ffmpeg, ffprobe, and Whisper. In a skill explicitly meant to fetch and analyze videos this behavior is expected, but it still exposes the host to network access, resource exhaustion, and potential exploitation of vulnerabilities in media-processing tools when handling untrusted content.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The script reads transcript content and formats it for submission to an external AI tool, but provides no privacy notice, consent gate, or warning that the transcript may contain sensitive data. In a video-analysis skill, transcripts can easily include private conversations, credentials spoken aloud, internal training content, or copyrighted material, so encouraging copy/paste or piping into a third-party service creates a real data-exfiltration risk.

Missing User Warnings

Low

Confidence: 89% confidence
Finding: The skill instructs users to download remote video content and generate multiple local artifacts, but it does not clearly warn about these side effects, storage impact, or trust implications of processing untrusted URLs. While not a direct exploit primitive, this omission can lead users to run the skill on attacker-controlled content or in sensitive environments without understanding that network access and local writes will occur.

Natural-Language Policy Violations

Low

Confidence: 96% confidence
Finding: The script's interactive output includes Turkish-only strings such as "URL gerekli!" and "Kullanım," and similar Turkish messages continue throughout the file. This imposes a specific language on users without opt-in or justification, which matches the language/locale policy violation category.

Natural-Language Policy Violations

Low

Confidence: 86% confidence
Finding: Comments and user-facing strings such as error messages and instructions are exclusively in Turkish. Under the policy, forcing a specific language without user opt-in can be a natural-language policy violation unless the locale restriction is clearly documented and justified.

Static analysis

No suspicious patterns detected.