spoken-response
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions include shell blocks that execute
agent-voiceby interpolating variables like$RESPONSE_AS_TEXTdirectly into command strings. This pattern can be vulnerable to command injection if the response content contains shell-active characters and the executing shell does not perform proper escaping. - [EXTERNAL_DOWNLOADS]: The skill requires the installation of the
agent-voicepackage from the author's public resources using theuvtool manager as part of the setup process. - [INDIRECT_PROMPT_INJECTION]: The skill processes text generated by the assistant, which could be influenced by untrusted conversation history, and uses that text as input for system-level commands.
- Ingestion points: Assistant response text (current or previous) is captured and assigned to variables in
SKILL.md. - Boundary markers: None are defined to separate the assistant's text from the surrounding command context.
- Capability inventory: The skill performs shell execution (
agent-voice speak) and local tool configuration (agent-voice setup) withinSKILL.md. - Sanitization: The instructions provide rules for formatting (e.g., converting HTML entities like
to normal spacing) but do not mandate specific shell-safe escaping for the interpolated content.
Audit Metadata