dalemoncayo/fullcase-web

README preview

# Fullcase Web

A Next.js 16 (App Router) + Firebase application used as a reference project for **demonstrating AI/LLM code review with Claude Code and OpenAI Codex**. The app itself is a lightweight test-case management tool (projects → modules → test cases → test runs); the interesting part of this repo is how it is reviewed.

---

## What this repo demonstrates

This repository is a working example of how to wire LLM-based code review into a real Next.js + Firebase codebase. It shows:

- **Project rules as context** — `CLAUDE.md`, `AGENTS.md`, and `GEMINI.md` at the repo root define the architecture, conventions, and review criteria that every AI reviewer reads before producing findings.
- **Custom subagents** — `.claude/agents/` and `.codex/agents/` define three specialized reviewers (quality, performance, security) that run in parallel.
- **Skills** — `.claude/skills/code-review/` contains the shared review checklist, rubric, and a `bad-review.md` of known false positives to suppress.
- **Slash commands / prompts** — `.claude/commands/review-pr.md` and `.codex/prompts/review-pr.md` define the review procedure both tools follow.
- **Tool permissions** — the Claude GitHub Action (`.github/workflows/claude-review.yml`) whitelists a narrow set of `gh` commands. The Codex Action (`.github/workflows/codex-review.yml`) runs with `sandbox: read-only` and `safety-strategy: drop-sudo`.
- **Triggered reviews** — commenting `@claude-review` or `@codex-review` on any PR runs the respective reviewer and posts a single consolidated comment back.

If you want to study the review wiring without running the app, start by reading, in this order:
1. [AGENTS.md](AGENTS.md) — canonical architecture + review rubric
2. [CLAUDE.md](CLAUDE.md) — Claude-specific guidance that extends `AGENTS.md`

View full repository on GitHub →