context-loader

Load project context efficiently by prioritizing manifests, conductor docs, and track specs to minimize data read. This approach respects ignore files, uses git-based discovery, and caps the loaded set to 15 files to reduce token usage.

Tiered context discovery prioritizes reading manifests, conductor docs, and track specs or plans over scanning the entire repository. It loads only essential project context files to enable faster decisions and reduce token usage.

Prepare project context for a new track by loading prioritized files such as CLAUDE.md, conductor/product.md, conductor/tech-stack.md, tracks.md, and the current track spec and plan. This orchestration gathers essential context for planning and execution.

Context loading respects ignore files during project structure evaluation. It uses git-based file discovery to ensure ignored files are skipped, directly contributing to token optimization and faster context retrieval.

The maximum number of files loaded for Conductor workflows is capped at 15 files. This file cap enforces token optimization and prevents excessive data reading when gathering project context.