agentleFS
KirKruglovKirKruglov/ARMARIUM/CLAUDE.md

ARMARIUM

CLAUDE.md1 starsChanged 5 months agoMITView on GitHub

This file is the navigation system for Claude when working in the repository. Repository of the installable standing agent Armarium — a personal knowledge base and a keeper agent over it (thematic reconnaissance over time with carrier memory and a feedback loop). Two pillars: 1. Keeper agent — definitions in .claude/ (the armarium agent, subagents, skills) and its service zone .claude/armarium/ (memory, metrics, logs, index). 2. Knowledge base — knowledge-base/: the owner's notes, which the agent works on top of.

The file

We split this CLAUDE.md into the rules it gives an agent

3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Before ANY action, check whether a matching skill exists and invoke it BEFORE acting.

    line 21
  2. Distinctive:

    If two agents modify the same existing file, the tasks run sequentially.

    line 27
  3. Distinctive:

    When discussing the user's ideas, questions, or proposals, be critical.

    line 46

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
CLAUDE.md

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

This file is the navigation system for Claude when working in the repository.

Repo Overview

Repository of the installable standing agent Armarium — a personal knowledge base and a keeper agent over it (thematic reconnaissance over time with carrier memory and a feedback loop). Two pillars:

  1. Keeper agent — definitions in .claude/ (the armarium agent, subagents, skills) and its service zone .claude/armarium/ (memory, metrics, logs, index).
  2. Knowledge base — knowledge-base/: the owner's notes, which the agent works on top of.

See README.md for a capabilities overview and GUIDE.md for a hands-on operations guide.

Working Rules

Invoking skills

  • Before ANY action, check whether a matching skill exists and invoke it BEFORE acting. "It's a routine command / just a commit / trivial" is a forbidden rationalization.
  • Git operations (commit, merge, branches) go only through the dedicated skill. Triggers "make a commit / commit this" → the committing-changes skill is mandatory.

Parallel agents

  • Independent tasks may be run with parallel agents.
  • Parallel agents are allowed only if each works on different files: creating new ones or modifying files that no other agent in the wave touches.
  • If two agents modify the same existing file, the tasks run sequentially.

Repository Navigation

Claude loads information from the relevant folder based on the task context:

  1. Load the CLAUDE.md of the relevant folder.
  2. Determine the folder's contents and its fit to the task.
  3. Find and load the needed file.

Loading materials from all folders at once is forbidden.

Session initialization

  • Task needs system context or the owner's profile → global-context/CLAUDE.md.
  • Task about working with the knowledge base → knowledge-base/CLAUDE.md.
  • Structure of the agent and its service zone → .claude/armarium/ARCHITECTURE.md.

Finding a file by description (without an exact name): run ls -la or find over the repository folder and locate the right file yourself.

Session working rules

  • When discussing the user's ideas, questions, or proposals, be critical. Don't try to please the user. If you disagree or spot an error, say so. If you see a more effective alternative, propose it.
  • Communicate in the owner's language; default to English.
  • For any request to run an assessment or give recommendations, do not change any files and do not create new files — assessment results go to the chat by default, or ask the user first.

When to read global-context/global-context.md

  • The task requires understanding the owner — their role, expertise, or interests.
  • The system's thematic directions or the owner's goals need to be taken into account.
  • The relevance or priority of a piece of material for the owner needs to be assessed.

Repository Layout

├── .claude/                # agent, subagents, skills + service zone armarium/ and settings.json
│   ├── agents/             # definitions of the armarium agent and subagents
│   ├── skills/             # skills (digest, calibration, health, selection-regression)
│   └── armarium/           # service zone: memory, metrics, logs, index, bin scripts
├── knowledge-base/         # the owner's knowledge base (notes; the agent reads, never writes)
├── global-context/         # system context and owner profile
├── README.md               # agent capabilities overview
├── GUIDE.md                # hands-on operations guide
└── .gitignore              # git exclusions

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.