content-discovery

Automate content discovery by configuring data sources and filtering rules, then use the /discover command to scan ArXiv, GitHub, and HuggingFace for relevant content based on your keywords. The Skill applies semantic filtering and deduplication to surface only high-quality, unique results matching your topics.

Yes. The Skill uses AI-driven semantic filtering with a 90% similarity threshold to deduplicate content across sources. It identifies and removes near-duplicate papers, projects, and models, ensuring your discovery results contain only distinct, relevant items.

After discovery completes, the Skill generates summaries and structured outputs with metadata and images, then publishes directly to your configured channels—Hexo blog, Telegram, or Discord. Multi-channel publishing happens in one workflow without manual steps.

Create discovery missions using /mission add to define your topics, data sources, and publishing channels via YAML/JSON rules. Once configured, execute /discover to run all enabled tasks. No external dependencies are required.

Yes. The Skill discovers new LLM models and datasets from HuggingFace filtered by your keyword rules and quality thresholds. It extracts metadata, applies semantic deduplication, and publishes findings to your chosen channels with provenance tracking.

Yes. The Skill generates AI-driven Chinese summaries alongside English metadata for all discovered content. Summaries and insights are published with full provenance metadata, supporting multilingual research workflows.