What this skill does

This skill is responsible for:

• Running a Python command from the repository root under conda run -n template_beginner.
• Ensuring the command is executed in the correct working directory (default: ``).
• Monitoring the subprocess and periodically reporting whether it is still running.

It does not interpret or modify any of the script arguments; it simply runs what the user provides.

When to use

Use this skill when the user wants to:

• Execute a Python script in the repo (e.g., python scripts/train.py ..., python scripts/forecast.py ..., or any other script).
• Run a custom Python script or helper where you want live progress and a clear exit status.
• Confirm a given python command completes successfully (including early failures).

How to run

1. Ensure you are at the repository root.
2. Run the python command using the template_beginner conda environment.

The canonical command is:

bash
1python .claude/skills/run-python/scripts/monitor_python.py \
2  --env template_beginner \
3  --cwd code \
4  --check-interval 10 \
5  -- python <your_script>.py <args...>

Note: conda run avoids needing to source activate and works in non-interactive shells.

Monitoring progress

This helper prints a heartbeat every few seconds while the python process is alive. It also exits with the same return code as the python process.

Example (run training script)

bash
1python .claude/skills/run-python/scripts/monitor_python.py \
2  --env template_beginner \
3  --cwd code \
4  --check-interval 10 \
5  -- python scripts/train.py model=shallow_water +experiments=shallow_water/det_mse n_epochs=1 device=cpu

Tips

• If conda run fails, confirm the template_beginner environment exists (see environment.yml).
• If the command hangs, inspect the script output to find exceptions and ensure any required data files exist.
• You can adjust --check-interval to print status more or less frequently.