The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on executing local Python scripts (scripts/dfs_client.py, scripts/schema_gen.py, and scripts/report_publish.py) to perform its primary functions. These scripts are invoked via the shell, which is a standard but sensitive capability.
[INDIRECT_PROMPT_INJECTION]: The skill architecture creates an attack surface by passing untrusted user inputs directly into command-line arguments.
Ingestion points: The skill takes search queries (<target query>), URLs (<hosting-page-url>), and file names (<report>.md) from the user.
Boundary markers: There are no instructions or boundary markers defined to prevent the agent from including shell metacharacters (e.g., semicolons, pipes) in these arguments.
Capability inventory: The skill uses python via a subprocess to run data collection and report generation scripts.
Sanitization: The instructions lack guidance on sanitizing user input before passing it to the script execution tools, relying entirely on the agent's internal safety logic to prevent command injection.

video-seo