Skill UI

Manage AI agent context by offloading data to files for just-in-time loading. Reduces context window bloat and enables dynamic discovery of relevant information. Supports scratch pads, plan persistence, and sub-agent communication to optimize long-term tasks and reduce token costs.

This skill guides the creation and maintenance of a `.geminiignore` file. By systematically filtering out machine-generated noise—such as build artifacts, dependency lock files, logs, and caches—it significantly optimizes the AI agent's context window. This process improves processing speed and, critically, reduces token consumption costs (FinOps).

This skill provides methodologies for designing sophisticated, layered memory architectures for AI agents. It moves beyond simple context windows and basic vector stores by incorporating short-term, long-term, and graph-based memory systems. Implementing this is crucial for agents that require persistence across sessions, maintaining complex entity consistency, and performing advanced, relationship-driven reasoning over accumulated knowledge bases, including time-aware fact tracking.

Expert in building Retrieval-Augmented Generation systems for LLM applications. Masters embedding models, vector databases, chunking strategies, and retrieval optimization. Capable of designing retrieval pipelines, implementing hybrid search, and optimizing context windows to ensure high-quality generation from retrieved documents.

This skill implements advanced 'Gatekeeper' logic to optimize AI agent performance in long-running workflows. It prevents context window bloat by sharding metadata, allocating token budgets, and enforcing Atomic Precision to deliver purely functional, filler-free responses. Ideal for complex coding, multi-step reasoning, and large document analysis where conciseness is critical.

This skill optimizes LLM context usage through prompt caching, log pruning, and AST-based file inspection. It enforces ultra-dense output, minimizes filler text, and manages token budgets efficiently. Ideal for technical tasks requiring precise code edits, structured data handling, and reduced context window consumption during long sessions.

This skill enables splitting extracted text into optimized chunks for LLM context windows or RAG ingestion pipelines. It supports configurable chunk sizes, overlap settings, and various chunker types including text, markdown, YAML, and semantic splitting. Users can utilize token-based sizing for specific models or standalone commands for pre-existing text. Ideal for preparing documents for vector stores or summarization tasks.