The Agent Skills Directory

[COMMAND_EXECUTION]: The skill automates the configuration of 'Stop' and 'Notification' hooks in the agent's settings.json file, enabling the execution of included scripts for status updates.\n- [SAFE]: The skill utilizes native macOS utilities (osascript, say, afplay) for local user feedback. Text-to-speech functionality leverages a well-known service via the edge-tts utility, representing a standard integration for speech synthesis.\n- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing environmental data. \n
Ingestion points: Metadata such as current working directory and git branch name are read from stdin in scripts/notification-hook.sh and scripts/stop-speak.sh.\n
Boundary markers: None present.\n
Capability inventory: Access to osascript for dialogs, tmux for pane switching, and audio playback.\n
Sanitization: While the scripts use jq to parse JSON, they lack specific escaping for AppleScript control characters when interpolating variables into UI alerts.

setup-notification