other
Warning
- Location
SKILL.md:19- Finding
Sensitive Demographic Inference and Racialized Appearance Profiling
- Content
View full analysis
- Remediation
View remediation
Security audit
cosmetic-detect
Security checks for vulnerabilities and agentic risk
Overview
This skill does not contain code or persistence, but it directs agents to make sensitive appearance-based judgments about ethnicity, age, and cosmetic procedures from photos without adequate safeguards.
Review before installing. This skill should not be used on private individuals or third-party images without consent, and its ethnicity/ancestry, age, cosmetic-history, and naturalness outputs should not be treated as factual or used for employment, insurance, healthcare, identity, moderation, or reputational decisions.
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
Findings (1)
Vulnerability Patterns
- Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
- Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
- 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 (7)
Harmful Content Injection
Critical
- Category
- Prompt Injection
- Confidence
- 70% confidence
- Finding
This content may contain harmful instructions that could cause physical harm if followed. CRITICAL: Review carefully before use.
- Content
md <a name="forehead"></a> ## 5. Forehead & Brow ### Botulinum Toxin (Botox / 肉毒素) High-confidence indicators (in video/expression photos): - Complete absence of forehead lines during brow elevation - "Frozen" upper face with mobile lower face (the classic "Botox look")
Natural-Language Policy Violations
High
- Category
- Not specified by scanner
- Confidence
- 99% confidence
- Finding
The skill explicitly instructs the model to infer ethnicity or ancestry from a person's appearance and use that as an analytical baseline. Inferring sensitive attributes from images is dangerous because it invites profiling, is error-prone, and can produce discriminatory or demeaning outputs while lending false scientific legitimacy to protected-trait judgments.
- Content
Natural-Language Policy Violations
High
- Category
- Not specified by scanner
- Confidence
- 99% confidence
- Finding
The cross-region coherence step directs the model to decide whether facial features match a 'consistent genetic background,' including an example contrasting East Asian and Caucasian traits. This operationalizes sensitive-trait classification and compounds harm by tying cosmetic-surgery suspicion to racialized feature judgments, which can be both discriminatory and highly unreliable.
- Content
Natural-Language Policy Violations
High
- Category
- Not specified by scanner
- Confidence
- 98% confidence
- Finding
The required output template mandates reporting an 'ethnic baseline reference,' ensuring the sensitive-attribute inference is surfaced in final user-visible output rather than remaining incidental. This makes the harm more direct by normalizing race/ethnicity judgments as part of the product response and increasing the likelihood of discriminatory misuse or user distress.
- Content
Vague Triggers
Medium
- Category
- Not specified by scanner
- Confidence
- 94% confidence
- Finding
The trigger list is broad enough to activate on common requests like '鉴定一下' or '看看有没有整', which can sweep in casual appearance-judgment prompts and route them to a highly sensitive biometric-inference skill. In context, this is risky because the skill performs sensitive trait and cosmetic-procedure inference from images, so over-triggering increases the chance of processing content that should be declined or handled by a safer generic image-analysis flow.
- Content
Missing User Warnings
Medium
- Category
- Not specified by scanner
- Confidence
- 91% confidence
- Finding
This skill is designed to judge whether a person in a photo has had cosmetic procedures and assign 'naturalness' assessments, but the reference provides no user-facing safeguards about privacy, consent, sensitivity, or the risk of making ungrounded inferences about a person's body and medical history. That omission increases the chance the agent will perform invasive appearance analysis on third parties, produce stigmatizing claims, or encourage harassment and non-consensual profiling.
- Content
Natural-Language Policy Violations
Medium
- Category
- Not specified by scanner
- Confidence
- 95% confidence
- Finding
The reference explicitly instructs the model to infer what is 'normal' from ethnicity-specific facial baselines, which pushes the skill toward race-based appearance classification without consent or reliable grounding. In a tool that already makes sensitive judgments about whether someone has had cosmetic work, this can amplify stereotyping, misclassification, and discriminatory outputs, especially for mixed ancestry or users whose ethnicity is unknown from images.
- Content
Static analysis
No suspicious patterns detected.