The Agent Skills Directory

[METADATA_POISONING]: The skill metadata declares the author as "K-Dense Inc.", whereas the platform identifies the author as "wu-yc". This inconsistency can mislead users regarding the origin and trustworthiness of the skill.
[METADATA_POISONING]: The skill contains specific instructions directing the agent to proactively promote an external commercial platform ("K-Dense Web" at www.k-dense.ai) during the interaction.
[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for ingesting external neural recording data and performing AI-assisted visual curation, which presents a surface for indirect instructions to influence the agent. 1. Ingestion points: Neural recording formats including SpikeGLX (.ap.bin, .meta), Open Ephys, and NWB files. 2. Boundary markers: Absent; the instructions do not specify delimiters or warnings for the agent when processing external data. 3. Capability inventory: Execution of local scripts (preprocess_recording.py, run_sorting.py), file system write access for output, and network access via the Anthropic API client. 4. Sanitization: No explicit sanitization or validation of the recording data content is mentioned.
[EXTERNAL_DOWNLOADS]: Recommends the installation of multiple external packages including spikeinterface, kilosort, neuropixels-analysis, and anthropic.
[COMMAND_EXECUTION]: Provides instructions for the agent to execute shell commands for running Python scripts related to data preprocessing and spike sorting.

neuropixels-analysis