piper
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill executes shell commands using user-supplied parameters for input and output paths. While these variables are quoted in the instructions, they represent a potential command injection surface if the inputs are not properly validated by the executing environment.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided text or markdown files to generate audio. This creates an attack surface where hidden instructions in the data could attempt to manipulate the agent.\n
- Ingestion points: The input parameter in SKILL.md accepts file paths or literal text.\n
- Boundary markers: None present in the piper execution steps.\n
- Capability inventory: File system read/write, local script execution, and binary execution.\n
- Sanitization: The clean_obsidian_for_tts.py script performs significant sanitization by removing markdown formatting, HTML, and non-ASCII characters.\n- [EXTERNAL_DOWNLOADS]: The skill instructions include downloading voice models from Hugging Face, which is a trusted source for model assets.
Audit Metadata