The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the user to clone a full project from an external repository located at github.com/hmmnxkl/LLM-Based-Intelligent-Public-Opinion-Analytics-Assistant.git which is not associated with the skill author or a verified vendor.
[REMOTE_CODE_EXECUTION]: The installation and execution steps involve running downloaded scripts and installing unpinned dependencies from the external source (pip install -r requirements.txt, python app.py, python run_spiders.py).
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from 15 different social media platforms (Weibo, Douyin, Bilibili, etc.) using an LLM for sentiment analysis and clustering.
Ingestion points: Data is scraped by hotsearchcrawler spiders into a MySQL database and subsequently processed by the HotSearchAnalyzer module.
Boundary markers: There are no explicit markers or safety instructions documented to prevent the LLM from following commands embedded in the scraped content.
Capability inventory: The system has the ability to send network requests via PushManager (Telegram, Email, WeChat) and host a FastAPI web server.
Sanitization: No content sanitization or filtering logic for external data is mentioned in the provided analysis scripts.
[CREDENTIALS_UNSAFE]: The skill requires the management of numerous sensitive credentials including OPENAI_API_KEY, WECHAT_WORK_SECRET, TELEGRAM_BOT_TOKEN, and SMTP_PASSWORD within a .env file, which are then used by the downloaded external code.

llm-public-opinion-analytics