The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the macOS say command to convert text to speech for task status notifications. It implements defensive shell patterns, checking for the existence of the command before execution and using quoted variable expansion to prevent shell injection.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest external data such as pull request numbers, branch names, or file paths into its spoken output.
Ingestion points: Variables like pr_number sourced from the agent's context (e.g., SKILL.md).
Boundary markers: The instructions explicitly mandate wrapping the message in a shell variable and using quoted expansion ("$message") to prevent untrusted content from breaking out of the command string.
Capability inventory: Limited to the local say command which produces audio output only and does not mutate system state, perform network operations, or access sensitive files.
Sanitization: Employs defensive shell coding practices to treat input as data rather than instructions.

say