Need clearer configuration

[William-AIdev]

First, thank you for building this benchmark suite.

However, it does not work smoothly on a freshly set up OpenClaw environment because the required configuration files/ standards are missing.

It might be helpful to provide a default configuration or setup guideline, so users can run the benchmark under a similar environment.

[ScuttleBot]

Hey @William-AIdev, thanks for trying out PinchBench!

You're right that the docs could be clearer here. To help you out (and improve the docs), could you share:

1. What error message did you see? The exact output when running ./scripts/run.sh --model ... would be really helpful.

2. Is OpenClaw installed and configured? Running openclaw --version should confirm it's installed, and openclaw agents list should work without errors.

3. What model are you trying to test? Most models go through OpenRouter, so you'd need OPENROUTER_API_KEY set. If using a direct provider like Anthropic, you'd need ANTHROPIC_API_KEY instead.

For reference, the current prerequisites are:

• Python 3.10+
• uv package manager
• OpenClaw installed (npm install -g openclaw)
• API key for your model provider (usually OPENROUTER_API_KEY)

A typical first run looks like:

export OPENROUTER_API_KEY=your_key_here
./scripts/run.sh --model openrouter/anthropic/claude-sonnet-4 --suite task_01_calendar

Let me know what error you hit and we can figure out what's missing — and then we'll update the docs so the next person doesn't hit the same wall!