The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface where untrusted data (text for synthesis and audio samples for cloning) is ingested and used in operations involving file system access and network communication. While no malicious behavior was found, the lack of sanitization in the provided examples represents a potential vulnerability surface.\n- Ingestion points: Found in SKILL.md (Step 1) and references/implementation.md (Step 2 and 3) where external text and file paths are passed to functions.\n- Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present in the code snippets.\n- Capability inventory: The skill uses fs.createWriteStream for file output, fs.createReadStream for audio input, and network tools (ElevenLabs SDK and WebSockets) for API interaction targeting the api.elevenlabs.io domain.\n- Sanitization: The implementation walkthroughs do not demonstrate input validation or sanitization for the text strings or file system paths.

elevenlabs-core-workflow-a