The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill reads external data from Apple Reminders (task titles and bodies), which serves as a potential vector for malicious instructions aimed at influencing the agent's behavior.
Ingestion points: Task properties such as name and body are retrieved via AppleScript as described in references/scripting.md.
Boundary markers: The instructions do not define specific delimiters or guardrails for isolating reminder content within the prompt context.
Capability inventory: The skill possesses the ability to execute AppleScript via subprocess calls and write operation receipts to the local file system.
Sanitization: The skill specifically mitigates command injection by requiring the use of AppleScript argv or temporary files for inputs, explicitly forbidding the interpolation of content into script or shell sources.
[COMMAND_EXECUTION]: The skill uses AppleScript to interact with the native Reminders application on macOS.
Evidence: The file references/scripting.md details the use of AppleScript commands such as make new reminder and instructions for executing these via subprocesses with on run argv to handle external data safely.

apple-reminders