The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions for installing well-known scientific Python packages (e.g., spikeinterface, kilosort, probeinterface, neo) from standard registries like PyPI and downloading pretrained models from the official SpikeInterface Hugging Face repository.
[REMOTE_CODE_EXECUTION]: The documentation mentions the use of containerized spike sorters (e.g., kilosort2_5) using Docker, which is a standard feature of the SpikeInterface library for cross-platform compatibility.
[DYNAMIC_EXECUTION]: The skill demonstrates the use of the UnitRefine model loader, which requires a trust_model=True flag to deserialize .skops files. The documentation includes explicit security warnings, advising users to only load models from trusted sources and to treat them like executable artifacts.
[CREDENTIALS_UNSAFE]: The skill correctly implements security best practices by instructing users to read API keys for LLM-assisted curation from environment variables (os.environ) rather than hardcoding them in scripts.

neuropixels-analysis