The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local PowerShell scripts using the -ExecutionPolicy Bypass flag. This allows scripts to run regardless of the system's execution policy, which is a common but potentially risky practice if scripts are modified or if the environment is untrusted.
[PROMPT_INJECTION]: The script scripts/parse_yt_studio.ps1 creates an Indirect Prompt Injection surface by parsing user-provided HTML files and saving the extracted text into markdown files which the agent is then instructed to read.
Ingestion points: The script scripts/parse_yt_studio.ps1 reads from an InputFile (e.g., topic.html) provided by the user.
Boundary markers: The generated markdown output (parsed_topics.md) does not use delimiters or instructions to treat the content as untrusted data.
Capability inventory: The skill utilizes subprocess execution for PowerShell and Node.js, and performs file system read/write operations across multiple directories.
Sanitization: There is no evidence of sanitization or validation of the text extracted via regex from the HTML file before it is written to the output markdown.
[COMMAND_EXECUTION]: The skill includes a 'Sync' command that uses Copy-Item to move files from a hardcoded absolute Windows path (C:\CreativeOS\...) to the user's profile. This facilitates the movement of potentially untrusted workflow instructions into the agent's active skill directory.

youtube-pipeline