The Agent Skills Directory

[COMMAND_EXECUTION]: The google_tts.py script uses subprocess.run to execute ffmpeg for concatenating audio segments. This is a legitimate functional requirement for generating audio for documents that exceed the API's single-request byte limit. The implementation correctly passes arguments as a list to prevent shell injection.
[PROMPT_INJECTION]: The skill reads and processes user-provided documents, creating an indirect prompt injection surface where instructions hidden in the document could influence agent behavior.
Ingestion points: extract.py reads and parses PDF, DOCX, and Markdown files.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are included in the prompt templates.
Capability inventory: The skill can execute system commands (ffmpeg) and perform network requests to Google Cloud APIs.
Sanitization: The clean_text function in extract.py removes markdown formatting, HTML tags, and URLs from the extracted text using regular expressions, providing a layer of defense against some injection patterns.

google-tts