The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on an external CLI tool named imsg to interact with the macOS Messages application.
[DATA_EXFILTRATION]: The skill provides commands to list chats, read message history, and monitor incoming messages (imsg chats, imsg history, imsg watch). This allows the agent to access and process sensitive private communications.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface where external data could influence agent behavior.
Ingestion points: Untrusted message content retrieved via imsg history and imsg watch commands in SKILL.md.
Boundary markers: Absent. There are no delimiters or instructions provided to the agent to treat message content as untrusted data.
Capability inventory: Includes the ability to send messages via imsg send (in SKILL.md) and perform other shell commands.
Sanitization: Absent. No filtering or sanitization of message content before it enters the prompt context.
[COMMAND_EXECUTION]: Operation requires 'Full Disk Access' and 'Automation' permissions on macOS, which are elevated system privileges that grant the skill significant access to user data and system control.

imsg