The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the sag binary to perform speech synthesis and generate audio files (e.g., /tmp/voice-reply.mp3).
[EXTERNAL_DOWNLOADS]: The skill installs the sag binary via a third-party Homebrew tap (steipete/tap/sag). This is a legitimate installation method for the specified utility.
[CREDENTIALS_UNSAFE]: The skill instructions and metadata properly utilize environment variables (ELEVENLABS_API_KEY and SAG_API_KEY) for authentication, following security best practices for secret management.
[DATA_EXFILTRATION]: The CLI tool communicates with ElevenLabs' official infrastructure to generate speech, which is the intended and primary function of the skill.
[PROMPT_INJECTION]: The skill interpolates user-provided text into shell commands for audio generation. While this presents a surface for indirect prompt injection, it is restricted to the text synthesis command and is considered low risk for this specific use case.
Ingestion points: User requests for voice replies in chat.
Boundary markers: None provided in the command templates.
Capability inventory: Shell execution of the sag utility.
Sanitization: Relies on the agent's internal handling of string interpolation.

sag