What problem does it solve? AI coding agents often waste tokens by loading entire repositories, dumping large terminal outputs, and repeating identical MCP calls, which inflates cost and degrades answer quality. This Skill enforces a disciplined context-efficiency layer that retrieves only the minimum information needed for a correct result. ## Core Features & Use Cases - Progressive Context Retrieval: Uses a funnel from existing context to symbol lookup, dependency expansion, semantic retrieval, and only then broader search, so the model never starts by reading the whole repository. - Output Compression and Externalization: Filters terminal noise (progress bars, ANSI codes, duplicate stack frames), summarizes large logs, and stores big artifacts externally instead of injecting them into context. - MCP Caching and Prompt Optimization: Normalizes and caches repeated MCP responses, compresses verbose instructions into compact constraints, and produces terse final responses without losing technical detail. - Use Case: When debugging a failing authentication middleware in a monorepo, the Skill identifies the affected package, retrieves only the target symbol, its callers, and relevant tests, then returns a concise root-cause explanation instead of scanning every file. ## Quick Start Ask the agent to fix a bug or explain a function in your repository and it will retrieve only the minimal relevant symbols and dependencies needed to answer correctly.
