ClawHub

Viral Video Analysis

Security checks across malware telemetry and agentic risk

Overview

This skill is purpose-aligned and disclosed, but users should know it sends video URLs to Memories.ai and produces partly template-based creator guidance.

Install only if you are comfortable sharing analyzed video URLs with Memories.ai and processing creator ROI spreadsheets locally. Use a dedicated API key, avoid private or restricted videos unless you have permission to submit them, and treat the PDF's coaching sections as template-guided recommendations rather than independent proof from the spreadsheet.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (7)

Tp4

High
Category
MCP Tool Poisoning
Confidence
84% confidence
Finding
The skill claims broad video-performance analysis using visual analysis and metrics, but the provided implementation mainly sends URLs to a third-party transcript API and evaluates word count, with additional spreadsheet processing beyond the headline description. This mismatch can mislead users about what data is processed and shared, increasing the risk of unintended external transmission and overbroad invocation in contexts where sensitive creator data or performance spreadsheets are involved.

Description-Behavior Mismatch

Medium
Confidence
95% confidence
Finding
The script presents itself as generating a report from the supplied Excel data, but large sections of the output are fixed conclusions, thresholds, coaching guidance, and reference links that are not computed from the dataset. In this skill context, that is dangerous because users may trust the PDF as evidence-based analysis for advertising spend and creator decisions, leading to misleading business recommendations and potentially harmful operational actions.

Intent-Code Divergence

Medium
Confidence
92% confidence
Finding
The module documentation claims the program generates a PDF report from video performance data, but the implementation embeds predetermined findings and recommendations regardless of input. This mismatch increases the risk of deceptive or overclaimed output, especially in a video-performance-analysis skill where consumers expect data-driven reporting rather than canned advice.

Vague Triggers

Medium
Confidence
80% confidence
Finding
Broad everyday invocation phrases increase the chance the skill is triggered in situations the user did not intend, especially when the skill can read files and send video URLs to an external service. In this context, accidental activation is more dangerous because the skill performs networked processing and may handle creator ROI spreadsheets, not just harmless local text transformations.

Vague Triggers

Medium
Confidence
82% confidence
Finding
The quick commands are open-ended and omit constraints around accepted inputs, data sensitivity, and external API use. Because some commands imply batch analysis and coaching over uploaded performance data, ambiguous triggering could cause unintended processing of spreadsheets or external sharing of video URLs without clear user awareness.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
User-supplied video URLs are transmitted to a third-party API without any clear runtime disclosure or consent mechanism. In an agent/skill context, this is privacy-relevant because URLs may embed identifiers, private links, or campaign information, and users may reasonably assume local-only analysis from the description.

External Transmission

Medium
Category
Data Exfiltration
Content
HEADERS = {"Authorization": API_KEY}

def get_transcript(url: str, platform: str = "instagram"):
    resp = requests.post(
        f"{BASE_URL}/{platform}/video/transcript",
        headers=HEADERS,
        json={"video_url": url, "channel": "rapid"},
Confidence
92% confidence
Finding
requests.post( f"{BASE_URL}/{platform}/video/transcript", headers=HEADERS, json=

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal