Agent Skills Hub

semantic-compression

Aggressively prune grammatical scaffolding and filler text from inputs to optimize LLM token usage while retaining core semantic content.

upstash-vector-db-skills

Upstash Vector DB setup, semantic search, namespaces, and embedding models. Ideal for building high-performance vector search features in Next.js 16/Vercel projects.

manufacturing-failure-reason-codebook-normalization

Normalizes testing defect logs by correcting typos, abbreviations, and ambiguous descriptions based on product-specific codebooks and station validation.

higress-auto-router

Manage automatic model routing for Higress AI Gateway via CLI. Configure triggers for intelligent model selection based on request content.

qmd

Local hybrid search engine for markdown notes, documentation, and codebase knowledge bases to reduce token consumption and improve retrieval efficiency.

taide-botrun

A RAG-based AI solver for high school Chinese GSAT exams, featuring structured knowledge retrieval, reasoning templates, and explainable AI outputs.

qras

A local RAG semantic memory system using Qdrant and Ollama. Ideal for recalling workspace files, notes, project decisions, and user preferences with high-relevance vector search.

fiftyone-find-duplicates

Find, review, and remove duplicate or near-duplicate images in FiftyOne datasets using computer vision similarity embeddings.

langgraph

Expert LangGraph architect skill for designing stateful, multi-actor AI agent workflows with robust persistence, conditional branching, and ReAct patterns.

ai-writing-detection

Comprehensive AI-generated text detection framework. Features multi-layer analysis of vocabulary, structural patterns, model-specific fingerprints, and technical metadata artifacts to identify AI authorship.

backend-rag-implementation

Build RAG systems to ground LLMs in proprietary data. Includes vector database integration, embedding strategies, hybrid search, and advanced retrieval patterns for FastAPI backends.

translate

Translate research papers (markdown) while preserving LaTeX formulas, code blocks, and images, with support for batch processing, retries, and portable bundles.