spoken-response-desktop

Pass

Audited by Gen Agent Trust Hub on Oct 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes the agent-voice command-line tool to convert text into speech. These commands are constructed by interpolating assistant response text into shell arguments in SKILL.md.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes text from assistant responses, which may contain content influenced by users, and uses it as input for shell-based tool execution.
  • Ingestion points: Assistant response text is captured in the RESPONSE_AS_TEXT variable in SKILL.md.
  • Boundary markers: There are no explicit markers or instructions to treat the interpolated text as non-executable data when passing it to the shell.
  • Capability inventory: The skill executes shell commands via the agent-voice utility.
  • Sanitization: Instructions are provided to normalize whitespace and literal escape sequences (like \n), but there are no requirements to escape shell metacharacters (e.g., backticks, dollar signs, or quotes) which could lead to command injection if malicious content is processed.
  • [EXTERNAL_DOWNLOADS]: The setup instructions in SKILL.md describe installing the agent-voice utility using uv tool install agent-voice and provide a link to the author's GitHub repository (https://github.com/yoav0gal/agent-voice).
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 17, 2026, 05:29 PM