smart personal fashion assistant

个人时尚助手 - 管理你的穿搭风格、电子衣橱，并提供每日穿搭推荐。支持身材肤色录入、衣橱增删改查、冲突检测、统计与智能推荐。

This skill appears internally consistent with a local wardrobe/fashion assistant, but review the following before installing: - Photo & data transmission: the tools compress/encode user photos into Data URIs and assemble payloads intended for a multimodal LLM. Confirm which model endpoint and credentials your agent will use — sending photos and personal body/skin data to an external model is sensitive. - Local filesystem writes: the skill will create ./smart_wardrobe/ (USER.md, inventory.md, wardrobe_data/, temp_upload/, wardrobe_dashboard/) and will move/delete files during import flows. Make sure you are comfortable with those files being created where the agent runs and that you have backups if needed. - Model-generated content becomes authoritative: several flows rely on model-generated IDs, HTML that contains embedded JSON, and then automatically update USER.md or inventory.md. A hallucinated or malformed model response could corrupt indexes or inject unexpected data into your local files — consider reviewing model outputs or enabling confirmation steps (the skill includes confirmation steps, but check your agent enforces them). - Execution capability: the bundle includes Python tools; the agent will only use them if it can run Python. If you run in a restricted environment, verify what runtime the agent will actually execute. If you want stronger safety: restrict the skill from autonomous invocation until you have validated end-to-end behavior, inspect the code in TOOLS/, and configure the agent so that any payloads containing images are sent only to a trusted model endpoint you control.

Type: OpenClaw Skill Name: personal-fashion-assistant Version: 1.0.1 The 'personal-fashion-assistant' skill bundle is a well-structured set of tools and instructions for managing a digital wardrobe and providing style recommendations. The Python scripts (e.g., wardrobe_tools.py, body_data_generator.py) handle local file management, image processing via PIL, and payload construction for LLM interactions without any signs of malicious intent or unauthorized data exfiltration. The Markdown files provide clear, functional instructions for the AI agent to interact with users and manage their fashion profiles (USER.md) and inventory (inventory.md). No evidence of prompt injection, obfuscation, or dangerous execution patterns was found.