researcher:init

Researcher Init

Initialize a research repository with the LLM Wiki structure for persistent knowledge accumulation.

Repository Structure

The initialized repository will have:

<repo>/
├── AGENTS.md          # Wiki schema and conventions (canonical)
├── CLAUDE.md          # -> AGENTS.md symlink
├── GEMINI.md          # -> AGENTS.md symlink
├── raw/               # Source documents (each in its own directory)
│   ├── <source_id_1>/
│   │   ├── content.md
│   │   ├── metadata.json
│   │   └── assets/    # Source-specific images/files
│   └── <source_id_2>/
│       ├── content.md
│       └── metadata.json
└── wiki/
    ├── index.md       # Content catalog
    ├── log.md         # Operation log
    ├── sources/       # Source summary pages
    │   └── <source_id>.md
    ├── entities/      # Entity pages (people, orgs, products)
    │   └── <name>.md
    └── concepts/      # Concept pages
        └── <name>.md

Note: Each source gets a unique directory raw/<source_id>/ where source_id is <sanitized_title>_<6_random_chars>.

Workflow

1. Check current directory:

   • If current directory is NOT a git repository → Initialize wiki structure in current directory (.)
   • If current directory IS a git repository → Ask user for repository path (default: ./research)

2. Run initialization script:

   python scripts/init_repo.py <path>
   

3. Confirm structure created and explain next steps

Path Selection Logic

Non-git directory (recommended for dedicated wiki repos):

• Initialize directly in current directory
• The entire directory becomes the wiki repository
• No subdirectory needed

Git repository (for project-embedded wikis):

• Create ./research subdirectory by default
• Keeps wiki separate from project code
• User can specify alternative path if needed

Next Steps

After initialization:

• Use /researcher:research <source> to ingest sources
• Use /researcher:consult <query> to query knowledge