smart personal fashion assistant

• Outcome: Pass
• Latest audit: May 28, 2026 · 4:51 AM UTC
• Version: 1.0.1

Overview

This is a coherent personal wardrobe assistant, but it handles sensitive body, style, and clothing-photo data that users should treat as private.

Install only if you are comfortable storing body measurements, skin/style attributes, wardrobe records, and clothing photos locally, and potentially sending some of that data to a multimodal model provider for analysis. Use a trusted model endpoint, review USER.md and smart_wardrobe contents periodically, keep backups before delete/update operations, and avoid directly running the Python tool files as scripts unless you understand their test code.

SkillSpector

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (16)

Missing User Warnings

Medium

• Confidence: 93% confidence
• Finding: The skill states that sensitive personal profile and preference data, including body measurements and style attributes, will be persisted to `USER.md` without a clear consent step or privacy disclosure. Because this is intimate personal data, silent persistence increases privacy risk, especially if the file is later accessed by other skills, synced, or exposed through logs/backups.

Missing User Warnings

Low

• Confidence: 84% confidence
• Finding: The installation flow includes automatic background initialization that creates files and directories, but the user is not clearly warned that filesystem changes will happen during setup. While the described behavior appears operational rather than malicious, undisclosed file creation reduces user awareness and can cause trust, integrity, or cleanup issues in shared or sensitive environments.

Missing User Warnings

Low

• Confidence: 80% confidence
• Finding: The skill advertises automatic daily and monthly proactive outputs based on stored wardrobe and profile data, but it does not clearly disclose this scheduled processing to the user. Even if the content is benign, recurring generation from retained personal data can surprise users and increase privacy concerns, especially when recommendations depend on lifestyle, location, or habit data.

Missing User Warnings

Medium

• Confidence: 93% confidence
• Finding: The code assembles outbound LLM payloads containing highly sensitive personal data: body measurements, skin-tone attributes, and base64-encoded user photos. In a fashion-assistant context, this is privacy-relevant biometric and appearance data, and the module provides no consent gate, minimization, redaction, or transmission warning before packaging it for an external model API, creating a real risk of unauthorized disclosure or non-compliant data handling.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The skill explicitly sends user-uploaded clothing images to an external multimodal model API for analysis, but the description does not warn users that their images leave the local system. This creates a real privacy and consent issue because uploaded photos may contain sensitive personal information, background details, or metadata, and users are not given an informed choice before transmission.

Missing User Warnings

Medium

• Confidence: 96% confidence
• Finding: The skill explicitly records persistent personal profile data in USER.md and uses wardrobe inventory plus clothing photos, but it does not clearly warn users that these data are stored and reused across sessions. This creates a privacy and consent problem: users may disclose sensitive body data, preferences, and images without understanding retention or downstream reuse.

Missing User Warnings

High

• Confidence: 98% confidence
• Finding: The skill instructs the agent to assemble user profile data, environment data, and wardrobe inventory into a prompt and send it to a large language model, but it never discloses this transfer to the user. Because the transmitted data can include body characteristics, preferences, schedule context, and linked clothing items/photos, undisclosed third-party/model processing materially increases privacy risk and may violate user expectations or policy requirements.

Vague Triggers

High

• Confidence: 94% confidence
• Finding: The trigger description allows activation not only from an explicit user request but also automatically on skill installation, which can initiate collection of sensitive biometric, body, preference, and photo data without a clear contemporaneous user action. In this context, broad auto-triggering is especially risky because the skill persists the collected data to USER.md and may prompt for image uploads, creating meaningful privacy and consent concerns.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: The skill is designed to collect sensitive personal data, including biological traits, body measurements, preferences, and photos, then persist both raw inputs and model-generated inferences into USER.md. Although it mentions a limited-use statement for photos later, it does not clearly warn up front that this information will be stored long-term, which undermines informed consent and increases privacy risk if the profile is later accessed or reused unexpectedly.

Missing User Warnings

Medium

• Confidence: 93% confidence
• Finding: The skill explicitly instructs the agent to send user fashion/profile data and photo paths to a multimodal model API, but it does not require informing the user that personal attributes and linked images may be transmitted to an external service. Because the data includes sensitive personal characteristics and potentially identifying photos, users may unknowingly consent to broader disclosure than expected, creating privacy and compliance risk.

Vague Triggers

Medium

• Confidence: 87% confidence
• Finding: The trigger description is broad and informal, so the skill may activate on loosely related requests and perform multi-step actions including reading profile/inventory files, calling a model, and writing outputs. Because the workflow also persists HTML and synchronizes data back into USER.md, ambiguous activation increases the chance of unintended file modification or disclosure during normal conversation.

Missing User Warnings

Medium

• Confidence: 95% confidence
• Finding: The skill explicitly instructs the agent to save model-produced HTML to disk and automatically sync extracted statistics into USER.md, but it does not require clear user notice or consent for those local file and profile modifications. This creates a real integrity and privacy risk: a user asking for analysis may not expect persistent storage, profile mutation, or acceptance of model-generated HTML that could later be opened in a browser.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The skill explicitly instructs the agent to send user clothing images to a multimodal model over the network for feature extraction, but it provides no user-facing disclosure, consent step, or data-handling warning. Because uploaded photos can contain sensitive personal information beyond the garment itself, silent external transmission creates a real privacy and data-governance risk.

Missing User Warnings

Medium

• Confidence: 90% confidence
• Finding: The skill explicitly instructs writing user preference data into USER.md but does not require an explicit warning, confirmation, or consent step at the point of modification. This creates a real integrity and privacy risk because persistent user-profile data can be altered or stored without sufficiently transparent user approval, especially when the skill is auto-triggered after profile creation.

Missing User Warnings

Medium

• Confidence: 89% confidence
• Finding: The skill explicitly retrieves and sends wardrobe images and 12-digit item IDs to the user, then reveals full item records after confirmation, but it contains no privacy guardrails, access-control assumptions, or minimization guidance. Personal wardrobe contents, images, and structured inventory metadata can be sensitive and may expose lifestyle, identity, location inferences, or other private information if surfaced to the wrong user, leaked in shared sessions, or over-disclosed.

Missing User Warnings

Medium

• Confidence: 93% confidence
• Finding: The skill explicitly allows irreversible deletion of wardrobe records and associated image files while making confirmation only optional. This creates a real risk of accidental or socially engineered data loss, especially because the skill edits persistent files (`inventory.md`, images, and logs) and the deleted item may not be easily recoverable.

Static analysis

No suspicious patterns detected.