video-add-content-cards

• Outcome: Pass
• Latest audit: Aug 10, 2026 · 5:58 AM UTC
• Version: 1.0.5

Overview

This skill is a coherent video-editing helper, but users should be aware that its preview workflow opens local HTML pages and some examples fetch third-party browser assets.

Install this if you are comfortable with a video-editing skill that reads and writes project working files, runs ffmpeg/Python/HyperFrames tooling, and opens local browser previews. In restricted or offline environments, review the example HTML first because some previews request GSAP and fonts from third-party CDNs; generated project review pages appear local and self-contained.

SkillSpector

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (7)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 88% confidence
• Finding: The skill clearly instructs the agent to read and write project files and execute shell commands, yet it declares no permissions. That mismatch creates a trust and enforcement gap: a runtime or reviewer may assume the skill is low-privilege while it actually performs filesystem and command execution actions, increasing the chance of unintended or unsandboxed execution.

Context-Inappropriate Capability

Medium

• Confidence: 81% confidence
• Finding: The skill mandates launching a local HTML file via native OS commands such as Start-Process, open, or xdg-open. Even though the target is a local example page, this expands the skill from media authoring into desktop process invocation and browser launch, which can be abused if file paths, associations, or local content are tampered with, and it is not strictly necessary for generating content cards.

Context-Inappropriate Capability

Medium

• Confidence: 84% confidence
• Finding: The review workflow again requires native OS commands to open generated HTML, giving the skill desktop-app control beyond simple plan generation and review-file creation. Requiring retries on failure further normalizes repeated process execution, increasing exposure to local environment manipulation, unsafe file associations, or unintended execution paths.

Context-Inappropriate Capability

Low

• Confidence: 92% confidence
• Finding: The file pulls executable JavaScript from cdnjs and fonts from Google at render time, introducing third-party supply-chain and privacy exposure into what is described as local content-card authoring. If those dependencies are unavailable, tampered with, or blocked, rendering can fail or load attacker-controlled code in the browser context.

Missing User Warnings

Low

• Confidence: 86% confidence
• Finding: This file loads third-party JavaScript and fonts from cdnjs, Google Fonts, and fonts.gstatic.com at runtime. That introduces a supply-chain and privacy exposure: external providers can observe requests, and if the CDN dependency is compromised or changed unexpectedly, the page will execute untrusted code in the rendering context.

Missing User Warnings

Low

• Confidence: 84% confidence
• Finding: The example silently makes outbound requests for a script and web fonts without any disclosure, which can surprise users in environments expecting local-only processing. This creates minor privacy and compliance risk because opening the file exposes client metadata to third parties and may violate offline or controlled-build assumptions.

Missing User Warnings

Low

• Confidence: 88% confidence
• Finding: The plan explicitly instructs opening generated local HTML pages via native OS commands (`Start-Process`, `open`, `xdg-open`) without any warning, confirmation, or trust boundary. Auto-opening generated artifacts can execute active content in the user's browser context and normalizes automatic execution behavior, which is risky when the HTML is derived from project data and may later incorporate unsafe content or external links.

Static analysis

No suspicious patterns detected.