create-speech-recording
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes the
agent-voiceCLI utility via shell commands to synthesize speech from text and to play back audio recordings. - [EXTERNAL_DOWNLOADS]: The skill provides instructions to install the
agent-voicetool using theuvinstaller and refers to the vendor's official GitHub repository for source code and documentation. - [INDIRECT_PROMPT_INJECTION]: The skill processes arbitrary text input and interpolates it into shell command arguments, which is an architectural attack surface for injection if inputs are not properly sanitized by the platform.
- Ingestion points: The skill ingests text content through the
$RESPONSE_AS_TEXTand$RESPONSE_AS_TEXT_FILEvariables inSKILL.md. - Boundary markers: No explicit delimiters or instructions are provided to the agent to isolate the input text from the shell command context.
- Capability inventory: The skill possesses the ability to execute shell commands, read files, and write to temporary system directories.
- Sanitization: While standard shell double-quoting is used in the command examples, the instructions do not include specific logic to sanitize or escape malicious shell metacharacters within the synthesized text.
Audit Metadata