context-gatherer

To extract codebase context before PRD generation, you can scan the repository to identify patterns, files, and entities from user prompts. This grounds feature specs by creating a structured context.md file under .claude/feature-state/{slug}/.

Codebase pattern detection for feature spec preparation is the process of extracting key terms like entities and domain words from prompts, then mapping them to existing project files. This reduces assumptions by embedding actual project context into specifications.

Yes, you can map prompt keywords to existing repository files. The skill extracts key terms such as actions and domain words from your prompts, searches the project to locate related files, and reads project context files to inform spec-building.

To generate a context.md file for feature state tracking, prompt the skill to scan your current project. It will extract terms, search the repository, read project context files, and output the report to .claude/feature-state/{slug}/context.md.

Yes, this codebase context extraction applies to any code project. It identifies patterns, files, and entities across the repository regardless of the specific tech stack, improving context accuracy before PRD generation.

Feature specs contain assumptions when they lack actual project context. You can fix this by scanning the codebase before PRD generation to extract key terms from prompts and map them to existing repository patterns and files.