Security audit

AI KEY·封面图制作

Security checks for vulnerabilities and agentic risk

Overview

This skill creates and saves video cover images with proportionate permissions, though photo privacy depends on the configured image provider.

Use an image provider you trust with the supplied photos; hosted processing may transmit both original images. Expect saved prompts and persistent design feedback. Review the separately supplied configuration, family rules, and related skills before use, since they are outside this package's inspection. State any language preference explicitly.

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

Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (3)

Reference photos are passed to image-generation tools without warning about potential third-party disclosure.

Medium

Category: Output Handling
Confidence: 90% confidence
Finding: The skill directs the agent to supply original person and background images to whichever compatible image tool is available, without checking whether that tool sends data to an approved external provider. If the tool is externally hosted, identifiable photos or confidential background details could leave the user's environment without informed authorization; local processing would not create this disclosure risk. Reference-image processing is necessary for this task, but that does not establish permission to disclose the images to any available provider.
Content: No source excerpt is available for this finding.

Platforms marked `domestic_public: true` require Chinese output without user language opt-in.

Medium

Category: Output Handling
Confidence: 92% confidence
Finding: The language rule directs the skill to produce Chinese versions for all platforms marked domestic_public: true, while offering language choice only for other platforms. The platform classification does not establish a requirement for Chinese or explicitly allow the user's preferred language to override it.
Content: No source excerpt is available for this finding.

The skill mandates Chinese-language Bilibili deliverables without user opt-in or a stated justification for the language restriction.

Low

Category: Output Handling
Confidence: 92% confidence
Finding: Line 25 requires both Bilibili covers to be in Chinese, without offering a language choice. The accompanying explanation justifies the aspect ratios and cropping requirements, not the mandatory output language.
Content: No source excerpt is available for this finding.