
Obsidian Kb
- Updated April 26, 2026
- szch79/agent-marketplace
Manages an Obsidian vault as a knowledge base: it ingests sources, distills insights from conversations, refines articles, and checks vault health. A developer or knowledge worker uses it to grow and maintain a clean, well-linked personal knowledge store.
Key points
- Obsidian vault management
- Ingest sources, distill insights
- Vault health checks
Obsidian Kb by the numbers
- Data as of Jul 7, 2026 (Skillselion catalog sync)
/plugin marketplace add szch79/agent-marketplace/plugin install obsidian-kb@my-claude-marketplaceAdd your badge
Show developers this skill is listed on Skillselion. Paste this into your README.
| Last updated | April 26, 2026 |
|---|---|
| Repository | szch79/agent-marketplace ↗ |
What it does
Manage an Obsidian vault knowledge base — ingest sources, distill conversation insights, refine articles, check vault health.
README.md
obsidian-kb
A plugin that turns an Obsidian vault into a progressively-refined knowledge base. An agent can ingest sources (PDFs, Zotero items, code repos), extract knowledge from conversations, and deepen articles by revisiting raw sources.
Skills
| Skill | Trigger | What it does |
|---|---|---|
kb-ingest |
"ingest this paper", "add this repo to KB" | Ingests a source into a curated KB article (see supported types below) |
kb-distill |
"distill the KB", "save what we learned" | Distills reusable knowledge from the current conversation |
kb-refine |
(model-invoked) | Deepens an existing KB article with more detail |
kb-search |
(model-invoked) | Teaches the agent how to search the KB via semantic similarity |
kb-health |
"check KB health", "run KB diagnostics" | Checks vault consistency (oversized notes, weakly-linked notes) |
Ingestion types
| Type | Example | How it's resolved |
|---|---|---|
| Zotero | darwiche02:knowledge_compilation_map |
Citekey → local PDF via Better BibTeX |
| PDF / file | ~/Documents/thesis.pdf |
Read directly |
| URL | https://react.dev/learn/hooks |
Available web fetch tool, follows same-domain subpages |
| Git repo | https://github.com/foo/bar |
Cloned locally |
Append a focus hint to scope extraction: /kb-ingest darwiche02:kc_map focus on DNNF and d-DNNF
Prerequisites
1. Obsidian vault
Create or designate an Obsidian vault for the KB. The plugin writes markdown files directly to this vault.
2. Environment variable
Set OBSIDIAN_KB_VAULT to the absolute path of your vault. Either in your
shell profile, Claude Code's settings (user level ~/.claude/settings.local.json or project level .claude/settings.local.json), or the equivalent Codex environment configuration:
{
"env": {
"OBSIDIAN_KB_VAULT": "/Users/you/Documents/MyVault"
}
}
3. Project-level permissions
The plugin invokes scripts via python3. Add this rule to your project's .claude/settings.json to avoid permission prompts:
{
"permissions": {
"allow": [
"Bash(python3 *obsidian-kb/*)"
]
}
}
4. Exclude asset files
Exclude ZOB_assets/ from Obsidian's file index so cloned repos don't pollute search:
Settings → Files and links → Advanced → Excluded files → add ZOB_assets/
5. Zotero (optional)
For Zotero ingestion, you need:
- Zotero running locally with the Better BibTeX plugin (for citekeys)
- The
resolve-citekey.pyscript queries Better BibTeX's JSON-RPC API directly
How it works
Articles are plain markdown files linked via [[wikilinks]]. Notes ingested from external sources have a citekey in frontmatter (e.g., citekey: path%/path/to/file.pdf) so the agent can resolve back to the raw source when needed. Structure emerges organically from wikilinks — discovery is through semantic search.
The lookup policy in CLAUDE.md teaches the agent to check the KB before reaching for raw sources or web search, and to flag gaps when knowledge is missing.