vercel-react-best-practices-report

A CLI/TUI tool powered by provider adapters to audit React and Next.js codebases against the $vercel-react-best-practices skill.

Installation

npm install -g vercel-react-best-practices-report

Or run directly with npx.

Prerequisites

Codex adapter

• A working Codex CLI environment on the machine.
• Valid Codex/OpenAI authentication for the current shell session.
• The $vercel-react-best-practices skill installed in the Codex-compatible skill location.

Claude adapter

• Claude Code CLI installed and authenticated in the current shell session.
• The $vercel-react-best-practices skill installed at ~/.claude/skills/vercel-react-best-practices/SKILL.md.

Usage

vercel-react-best-practices-report --adapter <codex|claude> --model <model-id> [--directory <path>] [--concurrency <n>] [provider options]

By default the tool runs with a single worker. Increase concurrency only when you explicitly pass --concurrency.

Options

| Option | Description | | --- | --- | | --adapter, -a | Required. Provider adapter: codex or claude | | --model, -m | Required. Model id for the selected adapter | | --directory, --dir, -d | Optional. Relative path under the current working directory to audit. Defaults to . | | --concurrency, -c | Optional. Number of parallel workers. Default is 1 | | --reasoning-effort, -r | Codex only. Passed through to Codex model_reasoning_effort. Defaults to high | | --effort, -e | Claude only. Passed through to Claude Agent SDK effort. Defaults to high |

Examples

vercel-react-best-practices-report --adapter codex --model gpt-5.4
vercel-react-best-practices-report --adapter codex --model gpt-5.4 --directory packages/web
vercel-react-best-practices-report --adapter codex --model gpt-5.3-codex --reasoning-effort xhigh
vercel-react-best-practices-report --adapter claude --model claude-sonnet-4-6
vercel-react-best-practices-report --adapter claude --model claude-opus-4-1 --effort max --concurrency 2

Output

Results are saved to react-best-practices-report/reports.json inside the audited directory. With the default --directory ., that means the current working directory.

The tool keeps the audit output contract fixed across providers: each file audit must resolve to a JSON object with a single findings array, and the host process validates and merges those results into the final report.

Visualize Results

Drag and drop reports.json into the viewer:

Vercel Best Practices Report Viewer