The Agent Skills Directory

[SAFE]: The skill relies on the rawgenai CLI tool, which is a vendor resource provided by the author (whq25) via a Homebrew tap. The tool acts as a wrapper for well-known AI services.
[COMMAND_EXECUTION]: The skill facilitates the execution of transcription commands through the rawgenai utility. These commands are well-documented and are the primary intended functionality of the skill.
[CREDENTIALS_UNSAFE]: Setup files (e.g., references/setup/openai.md) provide instructions on how to configure API keys for different providers using the tool's configuration command. No actual secrets are hardcoded; the skill follows standard security practices by using environment variables and configuration management.
[EXTERNAL_DOWNLOADS]: The DashScope provider reference mentions support for transcribing remote audio files via URLs. This is a standard feature of the Alibaba DashScope STT service and is documented as a native capability of the tool.
[PROMPT_INJECTION]: The skill provides clear boundaries and instructions for the agent. While the agent processes untrusted audio/video content (which could theoretically contain spoken instructions), the skill's guidelines focus on displaying the resulting JSON transcription to the user rather than executing the content as instructions.

agent-listen