Reference template for efficient AI agent documentation access (github.com)

This project is a reference-template repository that helps teams maintain local, version-controlled copies of frequently used technical documentation, optimized for AI coding assistants. By keeping docs locally indexed and scripted for automated updates, AI agents can read environment context and indexes (ENVIRONMENT.md and INDEX.csv) at session start, enabling faster, more accurate responses without web queries or external API calls. That reduces latency, network dependency, and token consumption while giving developers control over versions, offline availability, and reproducible agent behavior. Technically, the repo uses CSV indexing for efficient human/agent navigation, Git shallow clones and optional sparse checkouts to minimize disk use, and utility scripts (update-references, update-repos, update-docs) to fetch git and web-based docs. Minimal runtime dependencies are Git (2.25+), Bash 5+, fd, ripgrep, and optional snag or Kagi FastGPT for web fetching. You customize ENVIRONMENT.md, AGENTS.md, ROLE.md and clone preferred docs into ~/reference, then run ./update-references (suitable for cron). Agents are instructed to read ENVIRONMENT.md, INDEX.csv and local AGENTS/PROJECT files, and can then search with rg/fd across the offline corpus. The template is adaptable per workflow, branchable for different contexts, and released under MPL 2.0.