neuropixels-analysis
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [SAFE] (SAFE): No malicious instructions, obfuscation, or persistence mechanisms were detected. The skill is a legitimate research tool.
- [EXTERNAL_DOWNLOADS] (LOW): The documentation (e.g., in SPIKE_SORTING.md and standard_workflow.md) suggests installing community-standard packages like 'kilosort', 'spykingcircus', and 'mountainsort5' via pip. It also refers to official SpikeInterface Docker/Singularity images for running sorters in isolated environments, which is a standard practice for reproducibility.
- [COMMAND_EXECUTION] (LOW): The pipeline involves executing spike sorting algorithms which are often wrapped as subprocess calls or run within containers (Docker/Singularity). This is a core functionality of the SpikeInterface library and the intended use of the skill.
- [PROMPT_INJECTION] (SAFE): In AI_CURATION.md, there are prompts for vision-language models to analyze waveforms (e.g., 'Analyze this unit and determine if it represents a fast-spiking interneuron'). These are domain-specific, legitimate instructions and do not contain patterns intended to bypass AI safety guards.
- [INDIRECT_PROMPT_INJECTION] (LOW): The skill processes neural data visualizations via external AI APIs (Anthropic, OpenAI, Google). While this creates an ingestion surface where an attacker could theoretically craft adversarial data patterns to influence the curation LLM, the risk is minimal given the structured nature of electrophysiology data and the low impact of a misclassification in this research context.
Audit Metadata