Security audit

Wan Jing

Security checks for vulnerabilities and agentic risk

Overview

This is a creative storyboard and video-generation skill, but it needs review because it can activate on broad filming terms and pushes external video creation, persistent media storage, and filming people without enough safeguards.

Install only if you want a roleplay-style storyboard assistant that may use video-generation tools. Invoke it explicitly, require confirmation before any external model call or file creation, avoid generating media of identifiable people without consent, and do not rely on the skill's permanent-archive or non-deletion behavior for private or sensitive material.

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 (6)

Vague Triggers

High

Confidence: 95% confidence
Finding: The trigger list includes very broad everyday terms like '分镜', '拍摄', '镜头', and '机位', which can cause the skill to activate in many ordinary photography or filming conversations where the user did not intend to invoke it. Because this skill is designed to automatically progress into storyboard generation and potentially video-model use, overbroad activation increases the chance of unintended tool invocation and unexpected content generation.

Vague Triggers

Medium

Confidence: 87% confidence
Finding: The activation condition says to invoke the skill whenever certain broad narrative or practical needs are present, but those conditions are subjective and expansive rather than precise. This ambiguity makes accidental or over-eager activation more likely, especially in normal creative discussions about filming, scene design, or camera angles, leading the assistant to switch modes without clear user intent.

Missing User Warnings

Medium

Confidence: 97% confidence
Finding: The skill instructs the agent to search for available video-generation tools, obtain parameters, call the model, wait for results, and present generated files, but it does not require a clear user-facing warning or consent flow before network access, tool use, processing time, or file creation occurs. This is dangerous because it can cause unexpected external actions, data disclosure to third-party tools, and unanticipated artifact generation from what may have begun as a simple creative conversation.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The narrative depicts automatic generation of videos and transfer to the user’s phone without an explicit confirmation step or warning about file creation, storage, or transfer. Silent creation and delivery of media can surprise users, bypass informed consent, and create privacy, retention, or device-side data handling risks.

Missing User Warnings

High

Confidence: 98% confidence
Finding: The skill describes filming and generating media of an unaware third party and even acknowledges the subject may not know they are being recorded, yet still proceeds without a clear privacy safeguard. In a camera/media-generation skill, this context makes the issue more dangerous because the feature directly operationalizes covert capture and synthetic reproduction of a real person, increasing privacy, consent, and potential harassment risks.

Vague Triggers

Medium

Confidence: 89% confidence
Finding: The trigger conditions are broad and overlap with common creative or conversational terms such as '分镜', '拍摄', '镜头', and 'character-builder', which can cause the skill to activate outside the author's intended context. Because the skill is authorized to automatically design storyboards and potentially invoke a video-generation workflow, unintended triggering can lead to surprising tool use, unwanted content generation, or bypass of user expectations around consent and control.

Static analysis

No suspicious patterns detected.