The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted text input or file content and passes it directly to shell commands, creating a potential injection surface.
Ingestion points: User-provided text strings and file contents are processed in SKILL.md via the say and curl commands.
Boundary markers: The provided examples do not use delimiters or explicit 'ignore' instructions to isolate user input from the shell command structure.
Capability inventory: The skill has access to the Bash tool (used for say, curl, ffmpeg, and sherpa-onnx-tts), Read, and Write tools.
Sanitization: There is no mention of input validation or escaping to prevent shell metacharacters in user text from altering command execution logic.
[EXTERNAL_DOWNLOADS]: The skill references and requires several external components to function fully.
It instructs the installation of the sherpa-onnx Python package via pip.
It points to external model files and documentation hosted by the k2-fsa project on GitHub Pages.
It makes network requests to the ElevenLabs API (api.elevenlabs.io) to perform high-quality voice generation.
[COMMAND_EXECUTION]: The skill's primary logic is implemented through shell commands executed via the Bash tool.
Utilizes the native macOS say utility for immediate audio feedback.
Employs ffmpeg for converting AIFF files to MP3 format.
Uses curl and jq for interacting with remote TTS APIs and parsing their responses.

text-to-speech