The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes macOS command-line tools like duti, defaults, and killall to configure system-wide browser protocols and manage the Velja application state via the shell and Python subprocesses.
[DYNAMIC_EXECUTION]: Python scripts in the skill use subprocess.run to execute system commands for lifecycle management of the application, such as terminating processes and relaunching the service to apply configuration changes.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes potentially untrusted URL strings to perform routing and sanitization tasks.
Ingestion points: Raw URL strings processed by the test_velja_routing.py script and defined within routing rule configuration files.
Boundary markers: No explicit boundary markers are used to isolate or distinguish untrusted URL data from instructions during processing.
Capability inventory: The skill has the capability to execute shell commands and modify local configuration files (.plist).
Sanitization: Implements custom regex-based logic to remove common tracking parameters from URLs.

velja