int-brain-lab/ibl-ai-agent/ibl-neuropixel

IBL Neuropixel

Use this skill when

• The task involves raw SpikeGLX .bin or .cbin files.
• The user asks about destriping, compression, chunked binary reads, or probe-level preprocessing before ALF/ONE analysis.
• The user asks how to use tooling from the int-brain-lab/ibl-neuropixel repository.

Do not use this skill when

• The question is already about analysis-ready spikes, clusters, trials, or ONE-loaded objects.
• The task is primarily a scientific comparison, statistical contrast, or report-writing problem.

Local references

• references/neuropixel_routing.md
• references/neuropixel_function_signatures.md
• ../SOURCES.md

Default policy

1. Decide whether the user needs raw-ephys preprocessing or analysis-ready loading.
2. If raw binary access is needed, prefer repo-native spikeglx and ibldsp helpers before custom NumPy I/O.
3. If the task is destriping, compression, or chunked reading, follow the patterns documented in the upstream repo README.
4. Keep raw preprocessing separate from downstream scientific metrics.
5. Hand off to ibl-load or ibl-analyze once the data are in an analysis-ready form.

Boundary with other skills

• ibl-neuropixel: raw file mechanics, destriping, compression, waveform extraction, probe preprocessing.
• ibl-load: session-level loading through ONE/ALF or Brain Wide Map helpers.
• ibl-analyze: scientific metrics, PSTHs, decoding, statistics, and interpretation.