Seek And Analyze Video

• Outcome: Review
• Latest audit: May 29, 2026, 9:23 PM
• Version:
  2.1.1

Overview

This skill is not clearly malicious, but it should be reviewed because it encourages persistent external storage of meeting and customer videos without enough privacy or retention guardrails.

Install only if you are comfortable sending selected videos, recordings, images, notes, and queries to Memories.ai. Do not use it for confidential meetings, customer interviews, regulated data, or third-party recordings unless you have permission and an approved retention/access policy. Prefer quick non-persistent analysis for one-off work, and review the external GitHub repository before running code with your API key.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• 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
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (9)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 84% confidence
• Finding: The skill instructs users to set and rely on an environment variable API key, which is a sensitive capability, but it does not declare permissions or place limits around secret handling. In an agent environment, undeclared env access weakens transparency and can let a skill read or depend on secrets without clear user consent or policy review.

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 95% confidence
• Finding: The skill advertises powerful video discovery, analysis, storage, and cross-video querying capabilities, but the finding indicates the implementation only checks an environment variable and returns placeholders. This mismatch can mislead users into trusting fabricated outputs as real analysis, creating integrity and decision-making risks even if it is not direct code execution.

Description-Behavior Mismatch

Low

• Confidence: 79% confidence
• Finding: The documentation exposes destructive delete operations that are not reflected in the skill's analysis-oriented description, which can mislead users and downstream agents about the skill's actual authority. Hidden or under-disclosed destructive capability increases the risk of unintended data loss if an agent invokes deletion without clear user awareness or confirmation.

Missing User Warnings

Medium

• Confidence: 88% confidence
• Finding: The skill describes importing, storing, tagging, and querying videos and memories, but gives no warning about retention, third-party processing, or sensitive content exposure. Users may upload proprietary, personal, or regulated material without understanding where data is stored, how long it persists, or who can access it.

Missing User Warnings

Medium

• Confidence: 91% confidence
• Finding: Encouraging extraction of meeting notes from recordings without warning about confidential discussions, personal data, or consent requirements is risky. Meeting recordings often contain sensitive business information, legal matters, or employee data, and sending them to an external service can create privacy, compliance, and contractual exposure.

Missing User Warnings

Medium

• Confidence: 88% confidence
• Finding: The import workflow describes persistent storage of video content but omits any warning that data will be retained and may be queried later, which can cause users to submit sensitive material without informed consent. In a skill handling meeting recordings, competitor research, and personal video libraries, undisclosed retention materially increases privacy and compliance risk.

Missing User Warnings

Medium

• Confidence: 91% confidence
• Finding: The meeting-recording workflow encourages uploading and analyzing recordings without warning about consent, confidentiality, or the presence of sensitive information such as internal discussions or personal data. Because meetings often contain regulated or confidential content, this omission makes accidental privacy violations and unauthorized processing substantially more likely in this skill context.

Missing User Warnings

Medium

• Confidence: 89% confidence
• Finding: The meeting workflow explicitly imports internal recordings, extracts decisions and action items, and stores summaries to memory, but provides no warning or guardrails around sensitive business information, personal data, or consent. In this skill context, that is a real privacy/security weakness because users may process confidential meetings and persist the output into a searchable knowledge base without considering retention, access control, or data-minimization risks.

Missing User Warnings

Medium

• Confidence: 91% confidence
• Finding: The customer research repository example encourages importing customer interviews, user testing sessions, and external feedback videos, then synthesizing pain points and storing research insights, without any privacy warning. Because these materials can contain customer identities, proprietary information, or regulated data, the omission increases the chance of unsafe ingestion and long-term storage of sensitive customer information.

VirusTotal

65/65 vendors flagged this skill as clean.