The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted incoming message content, creating an attack surface where an external sender could attempt to manipulate the agent's behavior via text messages.
Ingestion points: Ingestion occurs when executing imsg history and imsg watch commands described in SKILL.md to retrieve message strings.
Boundary markers: Absent. The prompt instructions do not provide explicit delimiters or safety guardrails to separate untrusted text from system instructions.
Capability inventory: The agent can execute commands locally and send outbound messages via imsg send to arbitrary addresses.
Sanitization: Absent. No text validation or prompt escaping mechanisms are defined to sanitize input data.
[COMMAND_EXECUTION]: The skill requires the execution of local shell commands using the imsg CLI tool to interact with the host system's database and messaging stack.

imessage