ARMARIUM / knowledge-base
CLAUDE.md1 starsChanged 5 months agoMITView on GitHub
The owner's knowledge base — a durable, curated store of what the owner has learned and deemed important. It is "what the owner already knows," against which the keeper agent armarium synthesizes new material (definition — .claude/agents/armarium.md). Below, "the keeper" = this agent. When the owner saves material into the base that came from the keeper's digest, they tag the note with the line/tag #from-curator. This separates two feedback-loop signals: - general taste — any save into the base (what…
In this repository6 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- global-context/CLAUDE.mdCLAUDE.md
- calibrationSkill
- digestSkill
- healthSkill
- selection-regressionSkill
The file
We split this CLAUDE.md into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is 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.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 9
No MCP needed — the base lives in this same repo and is read directly.
- Very specific: line 13
When the owner saves material into the base that came from the keeper's digest, they tag the note with the line/tag #from-curator.
≈ “Common” is an estimate: at least 1 in 5 of the 27,641 CLAUDE.md 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
CLAUDE.md for the knowledge-base folder
The owner's knowledge base — a durable, curated store of what the owner has learned and deemed important. It is "what the owner already knows," against which the keeper agent armarium synthesizes new material (definition — .claude/agents/armarium.md). Below, "the keeper" = this agent.
Ownership and access
- Only the owner writes. The keeper is read-only: it reads the base with file tools and never writes here (exception — maintenance mode under per-item owner approval, where edits are marked as agent-authored).
- Base access for the keeper is a requirement, not an option.
- No MCP needed — the base lives in this same repo and is read directly.
The #from-curator marker convention
When the owner saves material into the base that came from the keeper's digest, they tag the note with the line/tag #from-curator.
This separates two feedback-loop signals:
- general taste — any save into the base (what the owner finds interesting at all);
- the keeper's own hit — a save tagged
#from-curator(what the keeper got right).
The marker name
#from-curatoris historical (the agent used to be calledcurator) and is kept as a stable signal interface — scripts and skills look for exactly this string. No need to change it.
Loop signal via git
- The folder is under git. The "what landed" signal is taken from the base diff between runs:
git diff <last_sha> -- knowledge-base/— scoped strictly to this folder, to isolate the signal from the keeper's own files in the same repo. - The keeper holds the last compared SHA in
.claude/armarium/memory/armarium-state.md. - Base commits are manual by the owner + a safety auto-commit of this folder by the keeper at the start of a run (catches anything uncommitted before the comparison).
Scope
- Domains are set by the owner in
.claude/armarium/sources.mdandglobal-context/global-context.md. - The storage structure inside the folder is arbitrary. The keeper only needs read access and a git diff over any structure.
Population status
Initially the folder is empty (only this CLAUDE.md). Move your notes here (for example, from Obsidian) — this is a manual owner step before the first end-to-end run. The denser the base on a topic, the better the synthesis: findings get linked to what has accumulated.
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.

