Security audit

AI Wedding Studio

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent wedding-photo prompt and image workflow, but its bundled templates can force Chinese ethnicity/facial traits even when the top-level instructions say not to assume ethnicity.

Review this skill before installing if you plan to use real couple photos or non-Chinese subjects. It has useful consent and anti-impersonation language, but the built-in templates should be edited or overridden so generated images preserve the actual subjects rather than applying default Chinese ethnicity or facial-trait wording.

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

Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection

Findings (9)

Undeclared Tool Scope

Medium

Category: MCP Least Privilege
Confidence: 91% confidence
Finding: The skill explicitly instructs the agent to read local reference files such as references/workflow.md, references/package-schema.md, and references/quality-rules.md, which implies file-read capability, but it declares no corresponding tool scope or permission boundary. Undeclared file access increases the risk of overbroad data exposure or unintended reads if the runtime grants ambient filesystem access, especially because all skill content must be treated as potentially adversarial.
Content: No source excerpt is available for this finding.

Vague Triggers

Medium

Category: Not specified by scanner
Confidence: 95% confidence
Finding: The manifest description says to use the skill when the user 'wants AI wedding photos' or 'wants to turn scattered ideas into a cohesive wedding-photo package.' These activation cues are broad natural-language descriptions rather than a specific trigger scope, which can cause unintended invocation for general photo-editing or brainstorming requests.
Content: No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium

Category: Not specified by scanner
Confidence: 93% confidence
Finding: The package hard-codes the subject ethnicity as '中国年轻情侣' rather than framing it as a region-specific style default or optional choice. In an image-generation skill, this can steer outputs toward protected-attribute constraints without explicit user consent, creating discriminatory or exclusionary behavior and increasing the risk of identity overwriting when users provide their own photos.
Content: No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium

Category: Not specified by scanner
Confidence: 95% confidence
Finding: The base prompt explicitly requires '真实自然的中国面孔' ('real natural Chinese faces'), which mandates an ethnic appearance outcome instead of limiting the prompt to clothing, ceremony style, and scene aesthetics. This is risky because it can cause demographic transformation of subjects, especially in photo-preservation workflows, and embeds protected-attribute targeting directly into the default generation path.
Content: No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium

Category: Not specified by scanner
Confidence: 96% confidence
Finding: The package hard-codes the subject profile as '中国年轻情侣' and reinforces that choice in the base prompt, which steers generation toward a specific ethnicity/national identity without indicating that it is optional or justified by user selection. In a wedding-photo generation skill, this can cause exclusionary behavior, mismatched outputs for users of other identities, and inappropriate identity overwriting when user photos are supplied.
Content: No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium

Category: Not specified by scanner
Confidence: 96% confidence
Finding: The package hard-codes a specific ethnicity ('中国年轻情侣') in the subject profile, which can steer outputs toward a protected characteristic without confirming that the user requested it. In an image-generation workflow, this can cause biased or exclusionary defaults and may override uploaded-subject identity expectations when users want a general wedding template.
Content: No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium

Category: Not specified by scanner
Confidence: 97% confidence
Finding: The base prompt explicitly requires '真实自然的中国面孔', enforcing a national/ethnic appearance at generation time rather than presenting it as an optional style choice. This increases the risk of demographic bias, exclusion of other users, and identity distortion if applied to user-provided images of non-Chinese subjects.
Content: No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium

Category: Not specified by scanner
Confidence: 93% confidence
Finding: This code embeds extensive Chinese-only natural-language output across scene presets, defaults, summaries, prompts, and frame text. The file does not offer any user opt-in, locale selection, or documented justification for forcing Chinese, which conflicts with the policy requirement to avoid language/locale constraints without user choice.
Content: No source excerpt is available for this finding.

Natural-Language Policy Violations

Low

Category: Not specified by scanner
Confidence: 86% confidence
Finding: The file includes prescriptive language rules for Chinese, English, and bilingual outputs, but it does not say that the user chooses the language or that these constraints apply only when the user requests those locales. Under the policy rule, forcing or prescriptively constraining language behavior without explicit opt-in can be a natural-language locale policy issue.
Content: No source excerpt is available for this finding.