The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes contact information which represents an untrusted ingestion surface. If a contact record contains malicious instructions, it could lead to indirect prompt injection when the agent reads the data.
Ingestion points: The scripts scripts/applescripts/contact/get.applescript and scripts/applescripts/contact/search.applescript read names, notes, and other fields from the Contacts application into the agent's context.
Boundary markers: The skill uses a structured JSON envelope and explicitly separates data fields, which helps the agent distinguish between command output and data.
Capability inventory: The skill possesses write capabilities including adding (add.sh), editing (edit.sh), and deleting (delete.sh) contact records.
Sanitization: String data is escaped using a custom jsonEscape function in scripts/applescripts/contact/_lib.applescript to ensure the integrity of the JSON structure.
[COMMAND_EXECUTION]: The skill executes osascript to perform macOS automation tasks.
Evidence: Files such as scripts/contacts.sh and AppleScript files use osascript and do shell script to interact with the system. These operations are essential for the skill's purpose and utilize proper sanitization, such as the quoted form of property in AppleScript, to prevent command injection from user-provided data.

macos-contacts