research-pack

To package a codebase for LLM analysis, provide a research topic and the skill dynamically selects the most pertinent files, bundling them into a ready-to-consume XML context pack under the .gdr directory for downstream deep research processing.

Smart codebase packing is the process of selecting the most relevant files from a codebase based on a specific research topic and assembling them into a compact XML context pack to reduce data overload and setup time for LLM analysis.

Yes, you can use this approach for codebases of varying sizes. It applies token budgeting to generate an under-1M-token XML pack, making it suitable for code review prep, topic exploration, and external LLM prompts without exceeding context limits.

You select pertinent files by providing a specific research topic to the skill. It then applies smart file selection logic to dynamically pick only the relevant files, ensuring the resulting XML context pack is focused and concise.

Yes, codebase packaging supports phase-based orchestration via agents. It coordinates deterministic tasks with parallel agents to efficiently select files and bundle them into the final XML context pack for deep research workflows.