The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill clones the Isaac-for-healthcare workflow repository from GitHub (https://github.com/isaac-for-healthcare/i4h-workflows) if the repository is not already present in the local environment. This repository contains the necessary scripts for the visualization workflow.
[COMMAND_EXECUTION]: The skill executes a shell script (viz.sh) from the cloned repository to serve the LeRobot HTML visualizer. It also uses git rev-parse to resolve project directories.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes metadata from dataset files, which represents a potential surface for indirect injection via data poisoning.
Ingestion points: Reads dataset properties from the meta/info.json file within the user-specified DATASET_DIR.
Boundary markers: The skill checks for file existence but does not implement specific boundary markers for the JSON content.
Capability inventory: The skill has the capability to execute bash scripts and start a local web server (typically http://127.0.0.1:9090/).
Sanitization: The skill verifies the dataset directory structure but does not explicitly sanitize the content of the info.json file.

i4h-lerobot-viz