The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes text from files or standard input to generate audio, creating a potential attack surface for malicious instructions embedded in the data.\n
Ingestion points: Text input via --text-file or stdin in the scripts/generate_voice.py script.\n
Boundary markers: No delimiters or safety instructions are provided to the agent to separate untrusted text from operational instructions.\n
Capability inventory: The skill performs network communication with the ElevenLabs API and writes binary data to the local file system.\n
Sanitization: Input text is transmitted to the external API without filtering or sanitization for instruction-like patterns.\n- [DATA_EXPOSURE_AND_EXFILTRATION]: The scripts/generate_voice.py script retrieves API credentials by searching for and reading .env files in the current and parent directories. It also accesses configuration from the user's home directory.\n- [PRIVILEGE_ESCALATION]: The helper script creates directories and writes files to paths determined by configuration or CLI arguments, allowing for local file system modification based on user-controlled input.

elevenlabs-tts