The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell scripts and AppleScript commands to interact with Contacts.app.
Uses osascript to run internal backend scripts from the scripts/applescripts/contact/ directory.
Employs do shell script within AppleScript for string normalization (e.g., in get.applescript), utilizing quoted form of to safely handle input strings.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from the user's address book.
Ingestion points: Contact fields (names, emails, notes) are read in scripts/applescripts/contact/get.applescript, scripts/applescripts/contact/list.applescript, and scripts/applescripts/contact/search.applescript.
Boundary markers: Absent; contact data is returned as plain strings within JSON fields without explicit delimiters for the agent.
Capability inventory: The agent can modify or delete user data via scripts/commands/contact/add.sh, scripts/commands/contact/edit.sh, and scripts/commands/contact/delete.sh.
Sanitization: Implements jsonEscape for output data formatting and quoted form of for internal shell command execution within AppleScript.
[SAFE]: No malicious patterns, such as hardcoded credentials, unauthorized network exfiltration, or obfuscated code, were detected. All external resource patterns trace back to the vendor's expected infrastructure.

macos-contacts