text-to-video

• Outcome: Pass
• Latest audit: Jul 25, 2026 · 4:39 PM UTC
• Version: 0.1.0

Overview

This skill is a disclosed text-to-video workflow whose shell commands, TTS providers, API keys, and output files fit its stated purpose, though users should treat cloud TTS as an external data transfer.

Install only if you are comfortable running local shell commands and using HyperFrames/FFmpeg tooling. Before selecting DashScope, OpenAI, or another cloud TTS provider, assume narration text may be sent to that provider; use macOS say or another local option for confidential drafts. Keep API keys in environment variables or a secrets manager, do not paste them into chat, and provide explicit input file paths rather than broad folders.

SkillSpector

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• 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: 90% confidence
• Finding: The skill instructs the agent to use shell commands and environment-based API key configuration, but the skill metadata shown in this file does not declare those capabilities explicitly. Hidden or undeclared execution and credential access increase the risk of overprivileged use, reduce auditability, and make it harder for users or policy layers to understand what the skill can actually do.

Vague Triggers

Medium

• Confidence: 86% confidence
• Finding: The quick-start examples encourage very broad free-form requests like converting arbitrary local files into videos without clearly constraining what inputs may be processed or when the skill should activate. In an agentic environment, ambiguous invocation patterns can lead to overbroad file access, accidental processing of sensitive local content, or unintended execution of downstream steps such as TTS generation and rendering.

Missing User Warnings

Medium

• Confidence: 87% confidence
• Finding: The skill asks users to provide text, URLs, and possibly scripts, then routes content to external TTS providers and third-party media sources without an explicit privacy warning. If users submit sensitive drafts, internal documents, or confidential URLs, that data may be transmitted to external services unintentionally.

Missing User Warnings

Medium

• Confidence: 93% confidence
• Finding: The skill tells users to configure API keys through environment variables but does not provide any credential-safety guidance. This can lead to accidental exposure through pasted keys, checked-in .env files, shell history, logs, screenshots, or generated project artifacts.

Missing User Warnings

Medium

• Confidence: 93% confidence
• Finding: The script sends arbitrary scene text from the input plan to DashScope's external TTS API, which is a real privacy and data-transfer risk if users include sensitive material. In this skill's context, the primary purpose is cloud TTS generation, so the behavior is expected rather than deceptive, but the lack of an explicit warning or opt-in means users may unknowingly transmit confidential text off-host.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The OpenAI path transmits scene text to an external TTS provider without an explicit warning, creating a genuine confidentiality and compliance concern for sensitive scripts. Because this skill is designed to convert user-authored scripts into audio/video deliverables, the risk is contextual and foreseeable, but still important since users may assume processing is local unless told otherwise.

Missing User Warnings

Medium

• Confidence: 91% confidence
• Finding: The template explicitly instructs users to configure third-party TTS providers and supply API keys, but it does not warn that scripts and potentially sensitive voice/content data will be transmitted to external services. In a text-to-video skill, users may paste proprietary scripts, internal documents, or personal audio content into the workflow, so omission of a disclosure and consent step creates a real data-handling/privacy risk rather than a purely theoretical issue.

Static analysis

No suspicious patterns detected.