agentleFS
devmesh-indevmesh-in/claude-company/CLAUDE.md

claude-company

CLAUDE.md8 starsChanged 3 months agoMITView on GitHub

Read this first, every spawn. This repo IS the claude-company product: the installer, hooks, skills, agents, and company/ canon that ship into user projects. When you work here you are building the machine other repos install, not running a company against an app. company/ ships verbatim into every install, so it must stay generic. The tracked company/gates.config keeps CONFIGURE ME placeholder gates on purpose - they are what a fresh install inherits before onboarding wires real commands. NEVER commit this…

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
  1. Very specific:

    Read this first, every spawn.

    line 3
  2. Distinctive:

    All six must be green before any commit, and CI runs all six plus the pack manifest, the readme check and the no-slop scan.

    line 25

≈ “Common” is an estimate: at least 1 in 5 of the 28,029 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 - project canon for claude-company

Read this first, every spawn. This repo IS the claude-company product: the installer, hooks, skills, agents, and company/ canon that ship into user projects. When you work here you are building the machine other repos install, not running a company against an app.

The dual-nature rule (read before touching company/)

company/ ships verbatim into every install, so it must stay generic. The tracked company/gates.config keeps CONFIGURE ME placeholder gates on purpose - they are what a fresh install inherits before onboarding wires real commands. NEVER commit this repo's real gate commands into it (BR-ASR-07). The suites that actually gate THIS repo are run directly, not through run-gates.sh:

python3 -m unittest discover -s tests/hooks -q   # the hooks
npm test                                          # CLI ONLY - tests/cli/test_cli.sh
bash tests/install/run_tests.sh                   # installer + the run-gates.sh runner
bash tests/install/test_tui.sh                    # the install TUI
bash tests/install/test_update.sh                 # update/rollout, incl. legacy field installs
bash tests/harness/run_tests.sh                   # the opencode adapter, renderer, drift gate

All six must be green before any commit, and CI runs all six plus the pack manifest, the readme check and the no-slop scan.

tests/harness/run_tests.sh REQUIRES the opencode binary on PATH and does not skip without it - a suite that skips itself when its subject is absent reports green for a machine where nothing was tested. Install opencode the same way you installed Claude Code.

npm test is the CLI suite alone. It does NOT run the installer, TUI or update suites - only prepublishOnly chains those. This file previously said two suites gate the repo and annotated npm test as covering install; both claims were wrong, and on 2026-08-13 that error reached three sealed briefs and cost a lane a full rebuild after CI caught a regression that two green local suites had missed. If you change anything under company/run-gates.sh, install.sh, update.sh or lib/, the installer suite is the one most likely to catch you, and it is the one nobody runs by habit.

Layout

PathWhat it is
bin/, lib/, install, install.shThe CLI and installer that copy the payload into a project
.claude/hooks/The enforcement hooks (Python) and their CLIs
.claude/skills/Slash commands (/company, /lean-company, /company-init, /brainstorm)
.claude/agents/The role definitions the CEO dispatches
.opencode/GENERATED view of .claude/ for the opencode harness, plus the hand-written adapter. Never hand-edit: run node lib/render-opencode.js
company/The canon, templates, and state that ship into installs
tests/tests/hooks/ (Python) and tests/cli, tests/install (bash)
docs/Client-facing guides

Enforcement facts

  • Hooks are Python 3.8 stdlib only and fail OPEN: an internal error lets the action through rather than jamming a session. The two integrity CLIs are the exception and fail LOUD - witness_check.py and trace_check.py on an orphan.
  • no_slop applies to ALL writing: straight quotes, ' - ' not em dashes, three dots not the ellipsis character, no stock AI filler. A CI job scans every tracked text file for the same glyphs.
  • The witness registry (company/witnesses.json) is checksum-sealed and mutated ONLY via python3 .claude/hooks/witness_check.py --add/--remove - never hand-edited.
  • Accepted ADRs (company/adr/, Status: accepted) are immutable: a guard blocks edits, and a settled decision changes only through a superseding ADR.
  • CLAUDE.md is NOT in the package.json pack list - it is repo-local canon and does not ship into installs. Keep it that way.

Commits

Conventional subject scoped to the workstream, a Task: <slug> trailer, and explicit staged paths - never git add -A. Work belongs on the task branch; the commit guard blocks direct commits on main while a task is active.

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.