Security audit

阿德影视制作大师

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed film-production writing assistant with local file/PDF output, and I found no hidden, destructive, credential-seeking, or exfiltration behavior.

Install only if you are comfortable with the skill processing pasted creative text and optionally writing scripts, prompts, or PDFs to local files. Be careful with confidential manuscripts or copyrighted source material, and decline PDF/export prompts when you do not want persistent local copies.

SkillSpector

By NVIDIA

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 (6)

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The README states the skill will auto-trigger when users mention broad categories like story adaptation, AI video production, or common tool names. This can cause the skill to activate during ordinary conversations that are only discussing those topics, unexpectedly steering the session into the skill’s workflow and increasing the chance of unintended processing of user content. In this context, the skill handles large user-supplied text and multi-stage transformations, so over-broad triggering is more dangerous than in a narrowly scoped utility skill.

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The activation description is very broad, triggering on common requests about story conversion, AI video production, or mentioning popular tools. This can cause the skill to activate in loosely related contexts and steer unrelated user conversations into a workflow that collects and transforms large amounts of user content, increasing the chance of unintended data handling or confusing task hijacking.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The skill instructs exporting user-provided creative content to PDF and saving current results, but it does not warn that this may create persistent local files containing sensitive or copyrighted material. In a system with file access, this can lead to unanticipated retention of personal data, confidential manuscripts, or transformed copyrighted text beyond the user's expectations.

Missing User Warnings

Low

Confidence: 81% confidence
Finding: The skill directs the agent to read an external local reference file during generation without warning the user that additional files may be accessed. Even if the file is benign, hidden local file reads expand the skill's data-access surface and can normalize behavior where agents consult local resources without clear disclosure or consent.

Vague Triggers

Medium

Confidence: 83% confidence
Finding: The trigger list contains broad, generic creative terms such as '影视制作', '影视剧本', and '短剧脚本', which can cause the skill to activate in many ordinary content-creation conversations beyond the user's clear intent. Overbroad activation increases the chance the skill is invoked unnecessarily with access to file read/write and PDF-generation capabilities, expanding exposure to prompt confusion or unintended processing of sensitive user content.

Vague Triggers

Low

Confidence: 74% confidence
Finding: The standalone trigger '即梦' is ambiguous and may match unrelated discussion, brand mentions, or casual text without a clear request to use this skill. This can lead to accidental invocation of a skill that handles user documents and generates outputs, increasing the risk of unintended data handling or workflow disruption.

VirusTotal

62/62 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.