The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill manages processes and monitors their output, creating a surface for indirect prompt injection if the supervised process generates untrusted logs.
Ingestion points: Process history and log scanning via zmx history as described in SKILL.md.
Boundary markers: The instructions do not specify delimiting process output from agent instructions.
Capability inventory: Arbitrary shell command execution via zmx run using bash -lc.
Sanitization: No explicit sanitization of process output is provided in the documentation.
[COMMAND_EXECUTION]: The skill is designed to execute shell commands to start and manage detached sessions. This is the primary intended purpose of the process supervisor and is not considered malicious.

zmx