elevenlabs
Pass
Audited by Gen Agent Trust Hub on Nov 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data sources to generate audio content.
- Ingestion points: The skill reads content from
VOICEOVER-SCRIPT.md,manifest.json, andproject.jsonto guide audio generation and scene synchronization. - Boundary markers: There are no explicit boundary markers or instructions to ignore embedded commands within the processed script files.
- Capability inventory: The skill has the ability to perform network requests via the ElevenLabs SDK, write files to the local system (e.g.,
save(audio, "voiceover.mp3")), and execute shell commands viauv runandffmpeg. - Sanitization: The instructions do not define sanitization or validation logic for the text input before it is passed to the ElevenLabs synthesis engine.
- [COMMAND_EXECUTION]: The skill documentation includes numerous examples of executing system commands to perform its tasks.
- Evidence: Use of
uv run tools/voiceover.py,ffmpegfor audio conversion, andcurlfor querying the ElevenLabs shared voice library. While these are primary functions of the skill, they constitute a command execution surface. - [EXTERNAL_DOWNLOADS]: The skill performs network operations to interact with ElevenLabs services.
- Evidence: The skill makes requests to
api.elevenlabs.iofor voice synthesis, transcription, and metadata retrieval. As this targets a well-known service for its intended purpose, these operations are documented as safe.
Audit Metadata