The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/generate.py performs local file system writes to save generated audio files based on user-provided paths.
[INDIRECT_PROMPT_INJECTION]: The skill accepts untrusted text and sends it to an external AI model for processing. Ingestion points: The text positional argument and the -f file input in scripts/generate.py. Boundary markers: Absent; user input is interpolated directly into the API prompt. Capability inventory: Performs network requests to Google and writes files to the local disk. Sanitization: Absent in the skill's code; the system relies on the Gemini API's internal safety filters for content moderation.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the google-genai package from a public registry, which is a legitimate dependency for Google services.
[DATA_EXFILTRATION]: The skill retrieves sensitive credentials from environment variables (GEMINI_API_KEY, GOOGLE_CLOUD_PROJECT). While these are used for authenticating with trusted services, the script includes a --no-ssl-verify flag that disables SSL certificate validation, potentially exposing these authentication keys and the generated audio data to Man-in-the-Middle (MitM) attacks if the feature is activated by the user.

gemini-tts