The Agent Skills Directory

[COMMAND_EXECUTION] (SAFE): The skill uses curl to interact with the ElevenLabs API. The use of bash -c is included as a specific workaround for environment-specific variable interpolation issues and is not used for malicious purposes.
[CREDENTIALS_UNSAFE] (SAFE): The skill follows best practices by using the vm0_secrets metadata to manage the ELEVENLABS_API_KEY. No hardcoded secrets were found.
[DATA_EXFILTRATION] (SAFE): Network activity is restricted to the official api.elevenlabs.io domain for the intended purpose of voice generation.
[INDIRECT_PROMPT_INJECTION] (LOW): The skill processes user-supplied text. It mitigates shell injection risks by writing the payload to a temporary JSON file and using the -d @file curl syntax instead of interpolating text directly into a command string.
Ingestion points: User-provided text in SKILL.md examples.
Boundary markers: Absent.
Capability inventory: Subprocess calls (curl), file-write (to /tmp/elevenlabs_request.json).
Sanitization: Uses JSON file payload for curl to avoid shell injection.

elevenlabs