The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill accepts user-provided input such as camera URLs and file paths that are directly interpolated into shell commands. This creates a potential surface for command injection if the input contains malicious shell metacharacters.
Ingestion points: camera-url and output-file arguments passed by the user as described in the frontmatter and examples.
Boundary markers: Not explicitly defined; the instructions use double quotes around variables in Bash command examples which provide some delimitation but not complete protection against sophisticated injection.
Capability: Uses the Bash tool to execute ffmpeg and grep commands.
Sanitization: The skill does not provide specific instructions for the agent to sanitize or validate the user-supplied strings before execution.
[SAFE]: The skill adopts safe practices by explicitly warning against logging camera credentials in output or memory. It uses established system utilities like ffmpeg and brew, and the suggested Python package onvif-zeep is a standard library for camera discovery.

camera