Claude Code Workflow Skills
A skill system for Claude Code that adds structured planning, real multi-model spec review, and session management. Four AI models review your specs in parallel — not simulated personas, but actual CLI calls to Codex, Gemini, and Cursor alongside Claude.
"Give Claude a way to verify its work." — Boris Cherny
This is the public, shareable skill system. It installs into your existing Claude Code setup alongside any personal configuration you already have.
About
Most AI coding workflows rely on a single model's perspective. This project brings genuine multi-model review to Claude Code — each model runs independently through its own CLI, then results are consolidated with consensus and divergence tagging. Combined with structured planning and session management, it turns Claude Code into a repeatable engineering workflow.
Features
- Multi-model spec review — Parallel review from Claude, Codex (OpenAI), Gemini (Google), and Cursor (Anysphere)
- Consensus tagging — Items flagged by 2+ models get highlighted; divergences are called out
- Planning wizard — Spec interview, review, scoping, and execution setup in one flow
- Session management — Start/end sessions with context preservation and automatic commits
- Knowledge capture — Document solutions for future lookup (first debug 30 min, future lookups seconds)
- Graceful fallback — Works with Claude alone if external CLIs aren't installed
Getting Started
Plugin (Recommended)
From any Claude Code session:
/plugin marketplace add matthewod11-stack/claude-setup
This installs the skills as a plugin — no manual file copying needed.
Legacy (install.sh)
git clone https://github.com/matthewod11-stack/claude-setup.git
cd claude-setup
./install.sh
Restart Claude Code after installing. The installer copies skills, reference docs, scripts, and templates into ~/.claude/ — it won't overwrite existing files in your solutions library.
Optional: External CLIs
For real multi-model reviews, install the external CLIs:
npm install -g @openai/codex @google/gemini-cli
# Plus: cursor-agent from cursor.sh
See Multi-Model Setup for details.
Skills
7 skills available via /command in Claude Code:
| Command | Purpose |
|---|
/plan-master | Master planning wizard with checkpoints |
/spec-review-multi | Real multi-model parallel spec review |
/roadmap-with-validation | Scoping + validation |
/session-start | Begin work session |
/session-end | End with commit + capture |
/checkpoint | Mid-session save |
/compound | Capture learnings |
Each skill is defined in skills/<name>/SKILL.md with YAML frontmatter specifying metadata and tool permissions.
Multi-Model Review
The /spec-review-multi skill launches real external AI CLIs for genuine review diversity:
| Model | Provider | Focus |
|---|
| Claude | Anthropic | Edge cases, security, architecture |
| Codex | OpenAI | Feasibility, API design, DX |
| Gemini | Google | Patterns, breadth, documentation |
| Cursor | Anysphere | File structure, modules, navigation |
Without CLIs installed, it falls back to Claude-only review. With CLIs installed, you get four genuinely different AI perspectives, consolidated with consensus and divergence tagging.
Tech Stack
| Technology | Role |
|---|
| Shell (Bash) | Multi-model orchestrator and CLI wrappers |
| Claude Code Skills | Slash commands and workflow definitions |
| Codex CLI | OpenAI review agent |
| Gemini CLI | Google review agent |
| Cursor Agent | Anysphere review agent |
Workflow
| Tier | Flow | Best For |
|---|
| Lite | Spec > Roadmap > Build | Side projects |
| Full | Spec > Review > Roadmap > Validate > Build | Production work |
Rule of thumb: could rebuild in a weekend? Lite. Otherwise, Full.
Documentation
Repo Structure
