Obsidian Vault Indexes Brand Data for Claude Content Generation

The Breakthrough

Duncan Rogoff created an Obsidian vault as a second brain for Claude Code, using five interconnected folders and index files to load personal story, audience language, proof, and content patterns, so Claude generates LinkedIn posts and Instagram scripts that sound authentic and convert without context rot.

What Actually Worked

• Vault divides into five folders: Audience (ICP profile, case studies, objections, proof bank, transformation arcs); Buildroom (courses, curriculum, projects, retention, testimonials); Coaching (raw transcripts from cohort, individual, community calls); Content (hook swipe files, what works on channels/competitors, video archive, email copy, platform specifics); AI (SOPs, commands, templates).
• Index file connects everything: links to offer positioning, brand voice, story hub, narrative arcs; subfiles branch to relevant data like coaching transcripts or curriculum for tasks.
• Core files include Offer & ICP (who served, transformation, exact words); Proof Bank (member wins, testimonials, before/after quotes); Brand Voice (sentence rhythm, what to avoid, how spoken, what works); Hooks (formats/angles driving DMs, channel-specific like YouTube/LinkedIn).
• Install Terminal plugin in Obsidian; run claude in terminal; paste Obsidian skills repo link and prompt Claude to install, enabling Obsidian Markdown skill for wiki links, embeds, properties, tags, aliases.
• Ingest data via Claude interview prompt: 'Interview me to collect everything needed then populate core files'; drop CSV testimonials or MD transcripts into raw folder and prompt 'ingest as proof' or 'extract high-level themes'; use Apify to scrape social posts (e.g., Instagram scraper with handle, export CSV) and prompt 'analyze performance to find what's working'; Claude desktop connects to Apify extension for direct scraping/analysis.

Context

Claude lacks memory, requiring re-explanation of self, offer, and audience each session, leading to context rot from overload. Rogoff's vault stores raw data (transcripts, social scrapes, testimonials) and distilled intelligence (hooks, ICP language), with graph view revealing connections; Claude loads only relevant nodes via index for tasks like post writing, pulling stories for credibility and CTAs to offers. This scales content creation for personal brands, mixing reach and conversion.

Notable Quotes

• 'The vault basically has you... everything about you your personal story any accomplishments... your proof anything about your audience your offer.'
• 'It is always going to sound like you because you have trained it on you and it's always going to speak directly to your audience because it's actually going to borrow their exact words.'
• 'Every expert who hits $10,000 a month and gets stuck has the same problem it's not their skill i've seen it across 70 plus coaching calls.'

Content References

No external books, papers, reports, podcasts, datasets, or events referenced beyond tools.