Vault Search Skill

Goal

Locate the most relevant notes in the vault using the obsidian CLI, rank them by relevance, and present actionable results with excerpts.

Prerequisites

pgrep -x Obsidian > /dev/null && echo "open" || echo "CLOSED"
# Read obsidian-cli skill before operating

⚠️ Never read or surface content from: dataScienceKnowledgeBase/dswok

Step 1 — Choose search strategy

For vague queries: run search:context first; if thin results, try search with synonyms.

Step 2 — Execute search

obsidian search query="<query>" limit=10

If empty results:

1. Try a broader synonym
2. Remove qualifiers
3. If still empty: "No notes found about X. Do you want me to create one?"

Fallback (if search times out — Obsidian in background):

find "." \
  -name "*.md" ! -path "*/dswok/*" \
  -exec grep -il "<query>" {} \; 2>/dev/null | head -10

Step 3 — Read and rank top results

obsidian read file="<note name>"

Read top 3–5. Rank by: title match → recency → note type (analysis > video > article).

Step 4 — Present results

📄 [[Note Title]]
Path: <RootFolder>/<folder>/<file>.md
Tags: #tag1 #tag2
Excerpt: "…relevant sentence from the note…"

Max 7 results. After listing, offer:

• Open in Obsidian: obsidian open file="<name>"
• Synthesize the found results → skill knowledge-synthesis
• Capture something new about this topic → skill note-capture

Constraints

• Excerpts only in initial results — full content only if user asks
• Never surface results from dswok
• If >20 results, ask for refinement before reading all
• Prefer obsidian search over grep/find

Examples

"What do I have on LangGraph?"

obsidian search query="LangGraph" limit=10
obsidian search:context query="LangGraph agents" limit=5

"Show me everything tagged #no-read-yet"

obsidian tag name="no-read-yet"

"Do I have notes on Claude Code?"

obsidian search query="Claude Code" limit=10
obsidian files folder="dataScienceKnowledgeBase/AI Engineer/Claude Code" total