Sources

Ground truth for the wiki. Files here are never edited or deleted. Claude reads them, extracts the intelligence into wiki/ articles, and moves each file from inbox/ to its classified folder.

To add material: drop it in inbox/ in any format (call transcripts, testimonials, research, notes, voice-memo transcriptions) and ask Claude to process it. After classification the file is renamed YYYY-MM-DD-type-slug, dated by when the source was created.

To pull Fathom recordings: set FATHOM_API_TOKEN in the project's ignored .env file and ask Claude to "sync Fathom". The sync follows every pagination cursor available to that key, imports only recordings not already filed, and then uses the same read-first classification rules. The token stays in .env and must never appear in a source, log entry, or command output.