ChiefOfStaff
CLAUDE.md4 starsChanged 8 months agoMITView on GitHub
ChiefOfStaff is a two-level autonomous multi-agent system for running businesses end-to-end. Level 1 (this repo): The "holding company" — 4 portfolio-level agents that evaluate ideas, design venture teams, create repos, and monitor the portfolio. Level 2 (venture repos): Each venture is a separate private GitHub repo with its own custom agent team, skills, commands, and scheduled tasks. There is no compiled code in this repo — all changes are markdown and YAML edits. Agents, skills, commands, and scheduled tasks are…
In this repository14 other files it gives its agents.
- brand-setupSkill
- daily-digestSkill
- deploySkill
- ecosystem-scoutSkill
- process-approvalsSkill
- publish-contentSkill
- qa-checkSkill
- send-outreachSkill
- submit-for-approvalSkill
- venture-bootstrapSkill
- venture-health-checkSkill
- venture-intakeSkill
- venture-staffingSkill
- weekly-reviewSkill
The file
We split this CLAUDE.md into the rules it gives an agent
9 rules. About 0 are common, 6 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 39
When scaffolding a new venture repo via skills/venture-bootstrap/SKILL.md:
- Distinctive: line 40
Always create a CLAUDE.md with venture-specific context, goals, and team
- Distinctive: line 41
Always create a README.md with venture overview and current status
- Very specific: line 42
Always create docs/ with architecture, requirements, market-research, business-model, changelog
- Distinctive: line 43
Always create custom agents tailored to the venture type
- Distinctive: line 50
Every change to agents, skills, commands, or scheduled tasks must be reflected in relevant docs
- Distinctive: line 57
Verify: Check that all cross-references resolve, all YAML is valid.
- Distinctive: line 58
Document: Update docs in the same change that modifies behavior.
- Very specific: line 59
Lessons: After any correction, update knowledge/lessons.md.
≈ “Common” is an estimate: at least 1 in 5 of the 26,980 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 — ChiefOfStaff Business Operating System
Project Context
ChiefOfStaff is a two-level autonomous multi-agent system for running businesses end-to-end.
Level 1 (this repo): The "holding company" — 4 portfolio-level agents that evaluate ideas, design venture teams, create repos, and monitor the portfolio. Level 2 (venture repos): Each venture is a separate private GitHub repo with its own custom agent team, skills, commands, and scheduled tasks.
There is no compiled code in this repo — all changes are markdown and YAML edits. Agents, skills, commands, and scheduled tasks are all markdown files with YAML frontmatter.
Key File Relationships
agents/*.md → skills/*/SKILL.md (agents are invoked by skills)
skills/*/SKILL.md → commands/*.md (skills are triggered by commands)
commands/*.md → scheduled/*.md (commands can be run by scheduled tasks)
portfolio/registry.md ← all venture state (lightweight index of all ventures)
templates/* → venture repos (templates are scaffolded into new repos)
docs/* → describes this system (must stay in sync with actual behavior)Agent Definitions
Agents are markdown files in agents/ with YAML frontmatter:
---
name: agent-name
description: When to use this agent
model: inherit
color: color-name
tools: [list, of, tools]
---Followed by role definition, principles, and behavioral rules.
Venture Repo Conventions
When scaffolding a new venture repo via skills/venture-bootstrap/SKILL.md:
- Always create a
CLAUDE.mdwith venture-specific context, goals, and team - Always create a
README.mdwith venture overview and current status - Always create
docs/with architecture, requirements, market-research, business-model, changelog - Always create custom agents tailored to the venture type
- Set up
.gitignore, initial commit, push to GitHub as private repo - Venture agents must update relevant docs when they make changes
Documentation Rules
docs/in this repo describes the ChiefOfStaff system itself- Every change to agents, skills, commands, or scheduled tasks must be reflected in relevant docs
docs/commands-reference.mdmust list all commands with current arguments and examplesdocs/architecture.mdmust reflect the actual system structure
Workflow
- Plan First: For changes touching 3+ files, write plan before starting.
- Verify: Check that all cross-references resolve, all YAML is valid.
- Document: Update docs in the same change that modifies behavior.
- Lessons: After any correction, update
knowledge/lessons.md.
Portfolio State
portfolio/registry.md— lightweight index of all ventures (slug, repo URL, stage, health)- Venture details live in their own repos, not here
portfolio/guardrails.md— ethical and spending rules that apply to ALL ventures
Scheduled Tasks
Scheduled tasks in scheduled/ define the 24/7 autonomous operation:
- Morning cycle (8 AM) — advance all ventures
- Evening digest (6 PM) — generate portfolio report → Notion
- Health check (every 6 hrs) — monitor venture health
- Weekly review (Friday 5 PM) — comprehensive portfolio review
Tools Available
- Notion — primary reporting dashboard (ventures database, digests, decisions)
- GitHub (
ghCLI) — create/manage venture repos - Claude CLI (
claude --task) — spawn Claude instances for venture work - WebSearch/WebFetch — market research, competitive analysis
- ContextPortal — institutional memory and context retrieval
- HubSpot CRM — contact/deal management for ventures in Launch/Growth
- Google Calendar — scheduling
- Eraser — diagrams and visual documentation
Notion Database IDs
These IDs are needed by agents when creating/updating Notion entries.
Your personal IDs should be in .claude/config.local.md (gitignored). Copy the template below and fill in your values:
# .claude/config.local.md
- **ChiefOfStaff page**: `YOUR_NOTION_PAGE_ID`
- **Ventures database**: `YOUR_VENTURES_DB_ID` (data source: `YOUR_VENTURES_DATASOURCE_ID`)
- **Digests database**: `YOUR_DIGESTS_DB_ID` (data source: `YOUR_DIGESTS_DATASOURCE_ID`)
- **Decisions database**: `YOUR_DECISIONS_DB_ID` (data source: `YOUR_DECISIONS_DATASOURCE_ID`)Also set environment variables or define in .claude/config.local.md:
$PROJECTS_DIR— where venture repos are cloned (e.g.,~/Projects)$CHIEFOFSTAFF_DIR— path to this repo- MCP server UUID mappings for
mcp__claude_ai_Notion__,mcp__claude_ai_HubSpot__,mcp__claude_ai_Miro__
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.

