price-check

搜索国内主流电商平台（淘宝/天猫、京东、拼多多、苏宁、唯品会、考拉、抖音、快手、1688）的实时价格，自动找到最合适的可信购买点并给出'值不值得买'建议 + 直接可点击的购买链接；本地积累历史价数据，能识别当前价是历史低位还是高位。Keywords: 比价, 值不值得买, 哪里买最便宜, 历史价, 价格监控, X...

This package appears to be what it says: a price‑comparison script that queries maishou88.com and keeps local history. Before installing/using: 1) Ensure your OpenClaw environment provides the Python dependency aiohttp (the script will fail otherwise). 2) Understand that all queries are written to ~/.openclaw/data/price-check/price-check.db (local storage) and config.json may store your Feishu token if you enable syncing — do not commit that file. 3) Feishu sync is opt‑in: only enable it if you trust the destination and you are comfortable storing/providing the base_token; enabling it runs lark-cli (a subprocess) which will execute a local binary. 4) The skill uses a default maishou API OPENID/invite code; if you have concerns about third‑party API usage, review or replace the data layer. 5) Note the SKILL.md rule that the agent must forward the generated human_report verbatim — be mindful if reports could contain sensitive text or links. If you want higher assurance, ask the author to declare Python package dependencies in the skill metadata and to document exactly what the default OPENID/invite code implies.

Type: OpenClaw Skill Name: price-check Version: 0.6.4 The price-check skill is a comprehensive tool for comparing product prices across major Chinese e-commerce platforms. It includes features for local price history tracking using SQLite (`db.py`) and optional synchronization with Feishu Bitables via `lark-cli` (`feishu_sync.py`). The core logic in `price_check.py` and `_data_layer.py` focuses on fetching data from the maishou88.com API, filtering for noise (like accessories or refurbished items), and calculating a 'verdict' on whether a product is worth buying. While the `SKILL.md` and JSON output contain 'Hard Rules' for the AI agent to render the report verbatim, these instructions are designed to ensure the integrity of the generated report and purchase links, rather than to facilitate an attack. No evidence of malicious intent, unauthorized data exfiltration, or harmful prompt injection was found.