The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to install several established scientific Python packages from standard registries, including spikeinterface, kilosort, and spykingcircus.
[COMMAND_EXECUTION]: Includes utility scripts and templates for orchestrating complex neural data processing pipelines. These scripts use standard argument parsing to manage data directories and processing parameters.
[DATA_EXPOSURE_AND_EXFILTRATION]: Features an AI-assisted curation workflow that optionally transmits waveform visualization data to major AI providers (Anthropic, OpenAI, Google) using official API clients. This is a core functional feature documented for unit quality assessment.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external scientific data files, which represents an ingestion surface.
Ingestion points: Raw neural recordings in .bin, .meta, .oebin, and .nwb formats are read via the spikeinterface and neo libraries in scripts/neuropixels_pipeline.py and scripts/preprocess_recording.py.
Boundary markers: Binary and structured data are processed through specialized scientific parsers that treat input as data rather than instructions.
Capability inventory: The skill can execute spike sorting binaries, perform disk I/O for saving results, and initiate network requests to AI APIs.
Sanitization: The skill relies on the validation and parsing logic of the underlying spikeinterface framework, which is standard for the electrophysiology field.

neuropixels-analysis