gdrive-knowledge
Skill0 starsChanged 5 months agoMITView on GitHub
Use when answering questions about Google Drive documents, files, folders, specs, contracts, meeting notes, proposals, or entities mentioned in documents. Drives the gdrive-knowledge MCP server tools (search, get_node, related_work, describe_schema, read_document).
In this repositoryOne other file it gives its agents.
- discord-searchSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 1 is distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 34
Call search(query) to find relevant nodes — it returns pointers + a small neighborhood, never content.
- Very specific: line 36
Call get_node(node_id, detail="full") to read the chunk text of a promising hit.
- Very specific: line 38
Call related_work(node_id) to explore folders, sibling docs, and entities around a hit.
- Very specific: line 39
Call describe_schema() if you need valid type/relation filter values before narrowing.
- Distinctive: line 45
If every gdrive-knowledge call returns an authorization error, your Google account is not linked to your identity yet.
- Very specific: line 48
Never invent document content.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Google Drive knowledge
Answer questions about your team's Google Drive using the gdrive-knowledge MCP server — a permission-filtered knowledge graph over your documents. Every result is masked before retrieval to exactly the documents you can read; forbidden documents are invisible, not "locked" — they never appear in results, neighborhoods, or counts.
When to use
Questions about documents, files, folders, specs, contracts, meeting notes, proposals, call transcripts, or entities mentioned in documents. Examples: "what does the consulting contract say about termination?", "find the design doc about embeddings", "which docs mention Northwind?".
Tools (gdrive-knowledge server)
| Tool | Purpose |
search | Search the knowledge graph. Returns node pointers plus a small neighborhood for top hits — never document content. Args: query, top_k, type_filter, relation_filter. |
get_node | Read a node. detail="full" returns the document's chunk text — this is the only tool that returns content. |
related_work | Explore neighbors (folders, sibling docs, entities) of a node you found. Args: node_id, hops (1–2), type_filter, relation_filter. |
describe_schema | Discover valid type_filter / relation_filter values for your allowed subgraph. |
read_document | Legacy alias for get_node(doc_id, detail="full"). |
Claude surfaces these prefixed by the server, e.g.
…gdrive-knowledge__search.
Workflow
- Call
search(query)to find relevant nodes — it returns pointers + a small neighborhood, never content. - Call
get_node(node_id, detail="full")to read the chunk text of a promising hit. This is the only content-returning tool. - Call
related_work(node_id)to explore folders, sibling docs, and entities around a hit. - Call
describe_schema()if you need valid type/relation filter values before narrowing.
Rules
- Results are masked to documents you can read. A missing document may mean "no access" — phrase as "I couldn't find or access it", never "it doesn't exist".
- If every gdrive-knowledge call returns an authorization error, your Google account is not linked to your identity yet. You cannot read Drive content until it is linked (see the connector's login flow in the README). Answer from Discord sources only in the meantime.
- Never invent document content; quote only chunks returned by
get_node.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

