The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions to execute shell commands including docker run for the primary mining task, nvidia-smi for hardware verification, and local Python scripts (prepare_nearest_neighbors_spec.py, verify_nearest_neighbors_spec.py) for configuration management.
[EXTERNAL_DOWNLOADS]: The skill pulls an official TAO Data Services container image from NVIDIA's container registry (nvcr.io). This is a legitimate vendor resource necessary for the skill's core functionality.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external data from parquet embedding files and YAML configuration specs which could potentially contain malicious content if sourced from an attacker.
- Ingestion points: source_parquet and target_parquet files, and the nearest_neighbors.yaml spec file.
- Boundary markers: The skill does not use specific delimiters but processes data through structured libraries.
- Capability inventory: Includes the ability to write result files and execute Docker commands.
- Sanitization: The provided Python scripts use yaml.safe_load() to parse configuration files securely.
[PRIVILEGE_ESCALATION]: The skill utilizes high-privilege Docker execution parameters, including --network=host, --gpus, and host-to-container volume mounting (-v). These flags are required for high-performance GPU-accelerated data processing and are standard practice for this specific toolset.

tao-mine-nearest-neighbors