uv inline script metadata for one-off matplotlib sims
To add a standalone matplotlib/numpy simulation folder to a repo without touching the app's dependencies, put a PEP 723 header (# /// script, dependencies = ["matplotlib", "numpy"]) at the top of the script and run it with uv run path/to/script.py; uv builds a throwaway env on the fly. Call matplotlib.use("Agg") before importing pyplot so it renders headless. For 'impact' Monte Carlo sims, keep every assumption as a named constant at the top and sanity-check the baseline first (my first baseline on-time rate came out at 55%, which wasn't believable for a team that finished most of its tasks). Also render each PNG and look at it: matplotlib legends and annotations often collide with the bottom row or a median line, so move them with loc= / ha='right'.