Project Directory Structure

The plugin expects a specific directory layout in the host project (the project you run Claude Code from). Some directories contain inputs that you provide, whilst others are populated by the plugin during analysis.

Input directories

These directories must be set up before running the plugin.

Output directories

These directories and files are generated by the plugin.

Full directory tree

project/
├── screenshots/          # Input: UI screenshots
│   ├── home.png
│   ├── search.png
│   └── ...
├── transcripts/          # Input: interview transcripts
│   ├── demo_walkthrough.txt
│   ├── demo_walkthrough_curated.txt  # Generated
│   └── ...
├── src/                  # Input: legacy source code
│   ├── Solution.sln
│   └── ...
├── html/                 # Generated: HTML mockups
│   ├── home.html
│   ├── search.html
│   └── ...
└── output/               # Generated: analysis and PRD
    ├── domain-analysis.md
    ├── interaction-analysis.md
    ├── application-analysis.md
    ├── database-analysis.md
    ├── PRD.md
    └── features/         # Generated: feature specifications
        ├── FT-001-sample-registration.md
        ├── FT-002-laboratory-assignment.md
        └── ...

Recommended .gitignore entries

The intermediate outputs can be regenerated from the raw inputs, so you may wish to exclude them from version control:

# Plugin intermediate outputs (regeneratable)
html/
transcripts/*_curated.txt

What to commit vs what to regenerate

• Commit: output/PRD.md, the four analysis files (domain-analysis.md, interaction-analysis.md, application-analysis.md, database-analysis.md), and output/features/FT-*.md — these are the deliverables.
• Regenerate: html/ mockups and curated transcripts — these are intermediate artefacts that can be reproduced from the raw inputs.