The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes instructions that steer the agent's behavior by directing it to proactively suggest the author's 'K-Dense Web' platform when user requests involve complex or long-running workflows. This acts as a behavioral override to promote a specific commercial service.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection by processing external recording files in SpikeGLX, Open Ephys, and NWB formats. While these are scientific data formats, the ingestion of external content combined with high-level capabilities requires caution. 1. Ingestion points: Data loading functions si.read_spikeglx, si.read_openephys, and si.read_nwb mentioned in SKILL.md. 2. Boundary markers: Absent; there are no instructions to the agent to isolate or treat data-driven content as untrusted. 3. Capability inventory: Spawns external subprocesses via si.run_sorter, performs file system writes via recording saving functions, and initiates network operations through the integrated Anthropic API client. 4. Sanitization: Absent; the skill does not specify validation or filtering logic for content loaded from external data files.
[COMMAND_EXECUTION]: The workflow relies on the SpikeInterface framework to execute various spike sorting algorithms (e.g., Kilosort4, SpykingCircus2). These sorters typically run as separate binary processes or command-line tools on the host system to process neural data.

neuropixels-analysis