The Agent Skills Directory

Unverifiable Dependencies (MEDIUM): The scripts scripts/setup-tensorboard.sh and scripts/setup-wandb.sh perform runtime installations of Python packages tensorboard, torch, and wandb using pip. While these are standard libraries, installing packages from external registries is a potential supply chain risk. Severity is reduced to LOW for the final verdict as these are essential to the skill's primary purpose.
Dynamic Execution (LOW): The setup scripts generate several Python files (e.g., example_usage.py, example_basic.py) using shell heredocs and apply chmod +x to make them executable. This follows the pattern of simple script generation from known templates.
Privilege Escalation (MEDIUM): The script scripts/launch-monitoring.sh uses pkill -f to terminate processes based on a user-provided port number. While arguments are quoted, interacting with process management via user-controllable strings requires caution.

monitoring-dashboard