Dicklesworthstone/agentic_coding_flywheel_setup

Agentic Coding Flywheel Setup (ACFS)

What it is – A Bash‑based installer that turns a brand‑new Ubuntu or Arch VPS into a ready‑to‑go “agentic coding” workstation. It pulls in a full development stack (shell, runtimes, cloud CLIs, etc.) and three AI‑coding agents (Claude Code, Codex CLI, Antigravity CLI) in a single, idempotent curl | bash command.

Why it matters – The repo bundles the entire toolchain needed for “AI‑assisted programming” workflows and makes the setup reproducible for individuals, teams, or classroom labs. By codifying the inventory in a manifest file and generating the installer scripts from it, the project aims for auditability and easy updates.

---

Core components (as described in the README)

Component	Path	Technology	Role
Manifest	acfs.manifest.yaml	YAML	Single source of truth – lists every tool, its install commands, and verification steps
Generator	packages/manifest/src/generate.ts	TypeScript + Bun + Zod	Parses the manifest, validates it, and emits Bash category libraries, doctor checks, and a harness
Website wizard	apps/web/	Next.js 16 + Tailwind 4	Interactive 13‑step guide that helps beginners rent a VPS, generate SSH keys, and run the installer
Installer	install.sh	Bash	Production one‑liner; sources the generated libraries and orchestrates the phased installation
Lib scripts	scripts/lib/	Bash	Reusable functions for security, checksum verification, health checks, updates, etc.
Generated scripts	scripts/generated/	Bash (source‑only)	Category‑specific install modules, doctor checks, manifest index, checksum ledger
Configs	acfs/	Shell/Tmux files	User‑level dotfiles (.zshrc, tmux.conf, onboarding lessons) deployed to ~/.acfs/
Checksums	checksums.yaml	YAML	SHA‑256 hashes for every upstream installer script; used by the security library

---

How the installer works (high‑level flow)

1. User runs the one‑liner (curl … | bash …).
2. install.sh sources the generated category libraries.
3. For each module (ordered by phase in the manifest) it runs the listed install commands as the required user (often root).
4. After a module finishes, the corresponding verify commands are executed; failures abort the run.
5. All external installers are fetched via HTTPS, their SHA‑256 hash is compared to checksums.yaml (via scripts/lib/security.sh), and only then are they executed.
6. Upon success the VPS ends up with:
   • Modern shell (zsh, oh‑my‑zsh, powerlevel10k)
   • Language runtimes (bun, uv/Python, Rust, Go)
   • Cloud CLIs (Vault, Wrangler, Supabase, Vercel, etc.)
   • Agent coordination tools (NTM, MCP Agent Mail, SLB, …)
   • Three AI coding agents ready to be invoked from the command line.

---

Key features highlighted in the README

• One‑liner, idempotent – Re‑run safely; it resumes from the last completed phase.
• “Vibe” mode – Pre‑configured for speed (password‑less sudo, aggressive agent flags).
• Wizard website – Guides absolute beginners through SSH key creation, VPS rental, and installer execution.
• Manifest‑driven generation – Adding a tool only requires editing acfs.manifest.yaml; the TypeScript generator updates all Bash modules automatically.
• Checksum‑based security – Every upstream script is verified against a stored SHA‑256 hash before execution.
• Reproducible team environments – Pin to a tag or commit (--ref) to guarantee identical installs across machines.

---

Typical use cases

Audience	What they get
Beginners	Step‑by‑step web wizard, a VPS that “just works” with AI agents ready to write code for them.
Individual developers	A single command that equips a fresh VM with a full stack and three coding agents, saving hours of manual setup.
Teams / classrooms	Pin‑to‑release installs ensure every member’s environment matches exactly, eliminating “works on my machine” problems for agentic workflows.

---

Getting started (from the README)

# Quick install (default to vibe mode)
{ acfs_installer="$(curl -fsSL \"https://raw.githubusercontent.com/Dicklesworthstone/agentic_coding_flywheel_setup/main/install.sh\")" \
  || acfs_installer="$(curl -fsSL \"https://cdn.jsdelivr.net/gh/Dicklesworthstone/agentic_coding_flywheel_setup@main/install.sh\")"; } \
  && printf '%s\n' "$acfs_installer" | bash -s -- --yes --mode vibe

For production you would replace main with a tag (v0.9.0) or a commit SHA and pass --ref so all scripts use the same version.

---

License & provenance

• License: MIT + OpenAI/Anthropic Rider (as indicated by the badge). The project therefore permits free use and modification while acknowledging the underlying AI‑model licenses.
• Platform support: Ubuntu 22.04 + or newer, and Arch‑based distributions.

---

Bottom line: ACFS is a concrete, open‑source bootstrapping tool that automates the creation of a VPS‑hosted AI‑coding environment. Its manifest‑driven architecture, checksum verification, and beginner‑friendly wizard make it a genuine contribution to the “agentic coding” niche.