Skip to content
Get packs

Vault Passage Retrieval

Search an Obsidian-style wiki vault for relevant passages using a local BM25 index with optional reranking.

Source Download skill
Repository
AgriciDaniel/claude-obsidian
Stars
16,065
Licence
MIT
Files
1 file

What it does

This skill teaches an agent to build a local search index over the wiki pages in a vault and use it to find the passages most relevant to a question. Search data is stored in a hidden .vault-meta folder and never changes the original notes. The agent can use a keyword-scoring method (BM25) by default, and optionally a semantic reranker running on a local Ollama model, falling back to keyword order if that is unavailable.

Great for

  • People who keep a large personal or team knowledge base in Obsidian-style markdown and want the agent to find relevant notes quickly
  • Developers tuning or evaluating retrieval quality on a vault with labelled test queries
  • Users who want search that stays on their own machine by default

Skip it if

  • Users who need the agent to send page text to a cloud API for prefix generation: this requires explicit consent, an API key and --allow-egress, and the data leaves the machine
  • Situations where the wiki is not a local folder of markdown pages, since the skill depends on the claude-obsidian wiki/ layout and its scripts
  • Anyone expecting search results to be proven facts: the skill states retrieval output is not evidence and the agent must still read and verify the pages

Try asking

  • “Find passages in my vault about BM25 and hybrid retrieval, top 5, read-only”
  • “Build the retrieval index for my vault using only local processing”
  • “Search my wiki for notes on project onboarding and explain which pages matched and why”
  • “Run retrieval diagnostics and benchmark the index against my labelled query file”
Step-by-step guide: how to use Vault Passage RetrievalInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 1 original files, the MIT licence and an attribution note.

    wiki-retrieve.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip wiki-retrieve.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip wiki-retrieve.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Start with --no-rerank and --no-llm so the first build and queries stay fully local and offline-friendly
  • Preview with --peek before any prefix build so you can see what data a provider would receive
  • Install the reranking model yourself beforehand; the tool will not pull it automatically

What's inside

  • SKILL.md

Redistributed unmodified from AgriciDaniel/claude-obsidian under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.