Back to search
ai-org-team
MarketplaceWorks with Claude CodeGitHub
Build and run an AI organization in Claude Code.
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Checked today
Badge for your README
Markdown
[](/listings/rishabhberia--ai-org-team)HTML
<a href="/listings/rishabhberia--ai-org-team"><img src="/api/listings/rishabhberia--ai-org-team/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
**As a plugin:**
/plugin marketplace add <owner>/ai-org-team /plugin install ai-org-team@ai-org-team
**Or as a plain skill:** copy `skills/ai-org-team/` into `~/.claude/skills/`.
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
.claude-plugin/marketplace.json · Manifest · 14 lines
- {
- "name": "ai-org-team",
- "owner": {
- "name": "ai-org-team contributors"
- },
- "plugins": [
- {
- "name": "ai-org-team",
- "source": "./",
- "description": "Build and run an AI organization in Claude Code: departmental AIs under a Management AI, talking through monitored sub-sessions."
- }
- ],
- "description": "Build and run an AI organization in Claude Code."
- }
README.md · README · 92 lines
- # ai-org-team
- A Claude Code skill that builds and runs an **AI organization**: one AI per department, each in its
- own repo with its own charter, knowledge base and memory, coordinated by a **Management AI** that
- acts as chief of staff.
- ```
- Organization AI team
- │
- Management AI ← routing, orchestration, conflicts, the ledger
- │
- ┌──────────┬─────────┼──────────┬────────────┐
- Sales AI Legal AI Finance AI Operations AI …
- ```
- ## How it works
- - **Each department owns its domain.** It is the only source of truth for it, and the only AI
- accountable for its numbers.
- - **AIs talk through sub-sessions.** When one AI needs information from another, or wants it to do
- something, it starts a Claude Code session in that department's folder. It watches that
- session, talks to it while it works, and records the request in a shared ledger. No AI reads
- another department's knowledge base to answer from it.
- - **Every answer carries provenance:** source, as-of date, basis, confidence, and what is missing.
- Missing is never written as zero.
- - **No AI approves anything.** Sending, writing to live systems, merging, deploying and spending
- stop with a named human principal.
- - **Confidentiality walls** hold in every direction, including toward the Management AI. These
- cover pay, personal data, privileged legal material, client-confidential terms and credentials.
- - **Conflicts are recorded, never averaged.** When two departments disagree, both positions go to
- the principal.
- The reasoning behind each rule is in
- [`skills/ai-org-team/reference/architecture.md`](skills/ai-org-team/reference/architecture.md).
- ## Requirements
- - **The Claude desktop app (Code tab)**, or any Claude Code harness that can start a session in
- another folder and message a running session. The protocol is built on sub-sessions.
- - `git` and `python3` (standard library only).
- ## Install
- **As a plugin:**
- ```
- /plugin marketplace add <owner>/ai-org-team
- /plugin install ai-org-team@ai-org-team
- ```
- **Or as a plain skill:** copy `skills/ai-org-team/` into `~/.claude/skills/`.
- ## Use
- Ask Claude something like:
- > Set up an AI team for my company with Sales, Legal, Finance and Operations.
- **You design the structure.** The skill asks what your organization does, proposes departments
- from a menu of common ones, and lets you add, drop, rename, merge, or split a department into
- several AIs (for example two Sales AIs, one per business line). It shows you the org chart
- (`scaffold.py --preview`) and changes it until you approve. Then it interviews you on each AI's
- scope, restrictions, systems and the confidentiality walls, and generates:
- | Repo | Contents |
- |---|---|
- | Management AI | `CLAUDE.md` charter; `knowledge/registry/` (department directory, capability matrix, systems map); `knowledge/protocol/` (interop protocol, data-sharing walls, escalation and authority); `knowledge/group/` (decision register, business context); `exchange/REGISTER.md` (the ledger); `logs/` |
- | Each department | `CLAUDE.md` charter with the team section; `knowledge/interop/ai-team.md` (team map and how to ask); `knowledge/00-INDEX.md`; `open-items.md`; `logs/` |
- Each AI's name, folder and knowledge-base folder are yours to choose. Existing projects can join
- as departments. Their `CLAUDE.md` is never overwritten. The skill
- writes a charter section for you to merge.
- It also covers adding, changing and retiring a department, changing the protocol across all
- departments, and a read-only health check (`scripts/check.py`).
- ### Optional modules
- - **Principal's office**: the Management AI also supports the principal directly. It handles
- briefs, follow-ups and drafts (never sent), and keeps personal matters walled off.
- - **PIAF** (Process Implementation & Assurance Framework): every new process or feature states its
- Promise, Proof and Protection, and production releases are checked against assurance gates.
- ## Try the scaffold without Claude
- ```bash
- python3 skills/ai-org-team/scripts/scaffold.py skills/ai-org-team/examples/example-config.json --preview
- ```
- ## License
- MIT
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.