From LoreDocs
Overview and orientation for LoreDocs. Use this skill when the user asks "what is LoreDocs", "how do I use LoreDocs", "how does LoreDocs work", "show me what LoreDocs can do", "tour LoreDocs", "intro to LoreDocs", or any other overview or orientation request about the product. Does NOT trigger on action phrases like "create a vault" or "search the docs" -- those go directly to the loredocs action skill.
How this skill is triggered — by the user, by Claude, or both
Slash command
/loredocs:using-loredocsThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
LoreDocs is a structured knowledge vault for documents that agents and humans
LoreDocs is a structured knowledge vault for documents that agents and humans produce over time -- architecture decisions, product specs, reports, guides, checklists. Unlike session logs (that is LoreConvo's territory), LoreDocs is for durable reference material that needs to be organized, versioned, tagged by type, and retrieved by topic.
You add a document once. Any future Claude session -- or any agent -- can search for it, inject it into context, or retrieve it by tag. Documents are versioned: every update keeps the full history, and you can roll back to any prior version.
Data lives locally in ~/.loredocs/loredocs.db and ~/.loredocs/vaults/. Nothing
leaves your machine.
Add a document when:
Inject context when:
LoreDocs ships these action skills:
| Skill | Trigger phrases | What it does |
|---|---|---|
loredocs | "create a vault", "store this document", "add this to the knowledge base", "search the docs", "inject context", "load project docs" | Core action skill: vault management, document add/search/retrieve, context injection, versioning. Start here for all operations. |
Use using-loredocs (this skill) for orientation. Use the loredocs action skill
for actual operations.
One vault per project. Create a vault named after the project and store all
project-specific docs there. Use vault_open_workspace("/path/to/project") to open
or create the workspace-scoped vault automatically.
Tag liberally. Tags drive cross-vault discovery (vault_inject_by_tag,
vault_search_by_tag). Tag every document with at least its topic area and category
(e.g., architecture, decision, spec). Tags are freeform strings -- no
controlled vocabulary required.
Use vault_inject_summary at session start, not vault_inject. A summary
loads titles and descriptions without full content, giving Claude a map of what is
available. Load specific documents on demand with vault_inject.
Categories and priorities. Categories (spec, guide, decision, reference,
checklist, report, other) and priorities (critical, high, normal, low)
are for your organization -- they are not enforced. Set them so agents can filter to
the most important material quickly.
MCP first, fallback second. LoreDocs works through MCP tools when available.
When the MCP server is unavailable, the bundled script (query_loredocs.py) does the
same job silently. You do not need to do anything special -- the loredocs action
skill handles the switch transparently.
DB lives at ~/.loredocs/loredocs.db -- not inside the project directory. If
you moved the plugin or changed the home directory, run loredocs-cli vault list
to verify the DB is accessible.
Free tier limits. Free: 3 vaults, 50 documents per vault, 1 MB per document.
Pro ($9/mo): unlimited. Check your usage with vault_tier_status().
Semantic search requires Pro. The vault_find_related embedding-based links
and vault_search with semantic=true require LoreDocs Pro. Free tier uses FTS5
keyword search, which is fast and covers most use cases.
First Pro install requires index build. After upgrading to Pro, run
vault_rebuild_index() once to build the LanceDB semantic search index. Subsequent
doc additions are indexed automatically.
FTS5 search tips. Use bare keywords for broad matches (architecture decisions).
Use double quotes for exact phrases ("session end"). Avoid special characters in
search queries.
Ask Claude to run:
vault_list()
If you see a list of vaults (or an empty list with no error), LoreDocs is connected
and working. If you get a tool-not-found error, the plugin is not loaded. Re-run
uv sync and reload the plugin.
Free tier covers 3 vaults with 50 docs each, FTS5 keyword search, full versioning,
and all context injection tools. Pro unlocks unlimited vaults and documents, semantic
search (embedding-based), automatic document relationship discovery, cross-product
session linking with LoreConvo, and the vault_rebuild_index tool.
Activate Pro by setting the LOREDOCS_PRO environment variable to your license key.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub labyrinth-analytics/claude-plugins --plugin loredocs