The Agent Skills Directory

[COMMAND_EXECUTION]: Executes high-privilege commands using sudo launchctl to manage system-wide services in /Library/LaunchDaemons and establish system persistence.
[COMMAND_EXECUTION]: Utilizes osascript to programmatically control user login items and executes an external binary (openclaw-gateway) as a persistent background process.
[CREDENTIALS_UNSAFE]: Provides instructions to store sensitive API keys (OPENCLAW_API_KEY) in plaintext within XML configuration (plist) files on the local filesystem.
[PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection by generating system-level configuration files from user-supplied data.
Ingestion points: Programmatic creation of .plist files from user-provided arguments.
Boundary markers: No specific boundary markers or 'ignore' instructions are implemented to separate user data from service logic.
Capability inventory: Full access to launchctl, sudo, and osascript for service lifecycle management and persistence.
Sanitization: Uses plutil -lint for syntax validation but lacks content-level sanitization to prevent malicious instruction injection into service parameters.

macos-launchd