Retrieval-Augmented Generation — search vault, build context, generate answers
This skill turns the Obsidian vault into a live knowledge base that Claude can query, reason over, and generate answers from. Every answer is grounded in YOUR notes, YOUR conversations, YOUR decisions.
Query ──→ RETRIEVE ──→ RANK ──→ AUGMENT ──→ GENERATE ──→ Answer + Sources
│ │ │ │
▼ ▼ ▼ ▼
Search vault Score by Build rich Generate answer
(Grep+Glob) relevance context citing [[sources]]
When given a question, expand it into multiple search vectors:
Execute these searches in parallel:
Score each retrieved note by:
Build the context window:
soul.md (identity grounding)Produce an answer that:
┌──────────────────────────────────────────────────┐
│ THE RAG LOOP │
│ │
│ 1. CAPTURE — Drop chats/files into Inbox/ │
│ │ │
│ ▼ │
│ 2. PROCESS — /project:process-inbox │
│ │ Extract knowledge → create notes │
│ ▼ │
│ 3. RETRIEVE — RAG skill searches vault │
│ │ Finds relevant knowledge │
│ ▼ │
│ 4. REASON — Claude analyzes and connects │
│ │ Generates insights + actions │
│ ▼ │
│ 5. ACT — Execute on insights │
│ │ Create business plans, make decisions │
│ ▼ │
│ 6. LOG — Daily digest captures what happened │
│ │ │
│ └──────→ Back to step 1 (new conversations) │
└──────────────────────────────────────────────────┘
"What do we know about meat export to the Middle East?" → Searches vault → finds StudEx Meat notes, StudEx Global Markets notes, any processed chats about export → synthesizes answer with [[sources]]
"What decisions have we made this week?" → Searches daily logs, notes with #decision tag, recent git commits → compiles decision log
"What's the most profitable opportunity right now?" → Pulls all business profiles, revenue streams, action items → ranks by ROI using business-analyst agent logic
The vault stays fresh through:
| Skill | How RAG Uses It |
|---|---|
| chat-processor | RAG searches notes created by chat processing |
| business-analyzer | RAG feeds data to the analyzer, analyzer feeds insights back |
| daily-digest | RAG includes daily logs in temporal searches |
| second-brain | Second Brain skill calls RAG for context when creating new notes |