Skip to content

Repository Template

Guide

Repository Template

{{PROJECT_NAME}} is simultaneously a live project and the canonical template source for new instances (ADR-0021). The scaffolding layer is the reusable payload; new repos start as a full copy of {{PROJECT_NAME}} and replace the {{PROJECT_NAME}}-specific content with their own.

Tooling on demand: agents are defined lean and tooled when the work demands it, via the access-request SOP (docs/sop/access-request.md). E.g. the Archivist is GitHub-tooled (ADR-0007); the PA is scoped to least privilege.

For a full setup walkthrough (including cloud services), see SETUP.md. This file is the background reference; SETUP.md is the action document.

What this is — an AI business ideator, incubator, builder & operator

This template is a framework for taking a venture from idea → operating company → continuous improvement. Every engagement starts with a needs analysis (GR-4) — who it’s for, the problem, constraints, and success criteria — then runs three roles. The template is the product; each repo instantiated from it inherits all of this:

  • Ideator — generate and sharpen the concept. (docs/concept.md, the operational-review prompt prompts/thoroughness-gaps.md, the Documentarian, and Golden Rule turning ideas into testable rules.)
  • Incubator — stand up the operating company around the idea: an org of purpose-built agents, HR hiring/onboarding, governance + policy, a preserved record, and one-command instantiation. (The org, hr/, archive/, the hooks, scripts/setup.sh.)
  • Mentor — provide ongoing, honest guidance: independent 360° reviews (External Review Partners), the daily briefing/report cadence, and the /agents:review | train lifecycle.

The Mentor guides every venture through the Standard Flow — Ideate → Plan → Build → Grow → Operate — with design research as the core spine. The flow is entered via a JSON ideation spec (from the monorepo) or direct chat — the only two entry points. See docs/methodology.md.

What you get (reusable scaffolding)

Carries over unchanged to every new repo:

  • Agent org — .claude/agents/ — 40+ agents across 19 units: Archivist, Historian, Golden Rule, Human Resources, Documentarian, Personal Assistant, Project Manager, Process Engineer, COO/CPTO/CGEO clusters, and more. Standard Claude Code subagents.
  • Lifecycle commands — .claude/commands/agents/ (/agents:hire, train, review, combine, fire) wired to HR.
  • Org chart & HR — ORG.md (units, roster, staffing model) + hr/ (employee handbook, policy library — security, resource-access, confidentiality, operating-cadence, ai-economy, … — hiring playbook, onboarding).
  • Golden Rules — CLAUDE.md (GR-1…GR-8: best model, CC workflows, co-create rules, analyze-before-tooling, two entry points, capture decisions, batch pushes, protect the archive).
  • Documentation system — docs/canon.md (macro→job-aid map), docs/adr/ (decisions), docs/sop/ (100+ SOPs), templates/ (doc templates), docs/knowledge/ (cumulative learnings), docs/glossary.md.
  • Preservation Archive — archive/ (charter, policy, MANIFEST.md, versioned documents/<doc>/{first,latest,milestones}/) with automated upkeep (scripts/archive-manifest.sh) and a CI gate (.github/workflows/ci.yml).
  • Intranet portal — intranet/ (Astro static portal over the repo: Dashboard, Docs, Chatter, People; deploys to Vercel behind single-user auth).
  • Operational-review prompt — prompts/thoroughness-gaps.md (generic; keep or adapt).

What’s project-specific (replaced by setup.sh)

scripts/setup.sh handles all of the below automatically:

  • Project identity — all {{PROJECT_NAME}} / {{project-slug}} references rewritten to your project name/slug
  • Principal identity — {{PRINCIPAL_NAME}} / {{PRINCIPAL_FIRST}} / {{github-user}} rewritten
  • GitHub URLs — github.com/{{github-user}}/{{github-repo}} rewritten to your username/repo
  • Vercel URLs — all *.vercel.app domains rewritten to your Vercel project names
  • docs/concept.md — not replaced by the script; you write this yourself (it’s the cornerstone)
  • history/chronicle.md — reset to a blank timeline by the script
  • reports/ + reviews/ — cleared by the script; yours will accrue
  • CHANGELOG.md — reset by the script
  • Archive — re-baselined by the Archivist agent after setup (manual step)
  • Cloud credentials — all new secrets; .env.example generated by the script

Create a new repository from this template

Fast path

# 1. Get a copy (GitHub "Use this template" button is the cleanest option)
git clone https://github.com/{{github-user}}/{{github-repo}}.git my-project
cd my-project

# 2. Run the setup wizard
bash scripts/setup.sh

# 3. Follow SETUP-STATUS.md for manual steps (GitHub, Vercel, Neon, Slack, archive)

See SETUP.md for the full walkthrough with every step explained.

What the wizard does vs. what stays manual

Automated by setup.sh Manual (see SETUP.md)
Identity substitution everywhere GitHub repo creation + branch protection
Directory renames Write docs/concept.md
intranet/package.json update Neon DB provisioning
Chronicle + reports + changelog reset Vercel project creation + auth
.env.example generation Slack app creation
SETUP-STATUS.md checklist Archive re-baseline (Archivist agent)

Invariants (carry to every repo)

  • Preservation discipline: first is immutable, latest is refreshed, everything is checksum-verified, and the Archivist receives everything.
  • Personal information is accessible only to the Personal Assistant and is never archived.
  • Unit Leads assume the work and hire only when it’s better allocated.
  • GR-7 (batch pushes): each push to a Vercel-connected branch consumes a deploy credit; commit freely, push at checkpoints.