ClawHub
Back to skill

Security audit

Ai Video Director

Security checks for vulnerabilities and agentic risk

Overview

The skill is a disclosed AI video-generation workflow; its main risks are expected third-party API use and some overly generic routing text, not hidden or malicious behavior.

Install only if you are comfortable with a video skill that can run local commands such as ffmpeg or ping and send prompts, scripts, product details, persona IDs, voice settings, and media metadata to third-party AI/TTS/video services. Avoid using sensitive unpublished business content unless you trust those providers and have configured API keys safely in environment variables.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill includes very broad, generic activation text such as instruction parsing and execution, which is not scoped to video-production tasks. In agents that auto-select skills from descriptions, this can cause the skill to trigger on unrelated user requests and gain access to read/exec capabilities in contexts the user did not intend. Because this skill advertises exec support, overbroad routing increases the risk of unintended command execution or unnecessary handling of sensitive user data.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill describes sending scripts, product information, agent identifiers, and possibly media-related metadata to external APIs, but it does not provide a clear user-facing warning or consent boundary about network transmission. Users may unknowingly submit sensitive business, marketing, or persona data to third-party TTS/video/LLM providers. In a content-generation skill with multiple external engines and fallback paths, this omission is more dangerous because data may traverse several services under failure conditions.

Content

No source excerpt is available for this finding.

Vague Triggers

Low
Category
Not specified by scanner
Confidence
91% confidence
Finding

The generic data-processing description makes the skill appear suitable for arbitrary transformation tasks instead of narrowly defined video workflows. This broadens the chance of accidental invocation for unrelated content-processing requests, which is especially problematic because the skill exposes exec and may interact with external APIs. The issue is contextual rather than directly exploitative, but it weakens containment and least-privilege expectations.

Content

No source excerpt is available for this finding.

Vague Triggers

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The result-output section is framed as a universal result validation/output capability without defining when the skill should activate. In auto-routing systems, such vague wording can make the skill match many benign requests, creating unnecessary exposure of its read/exec capabilities and increasing the chance of accidental execution in the wrong context. The surrounding skill is a media-production tool, so generic output language is unnecessarily risky.

Content

No source excerpt is available for this finding.