The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to facilitate interaction with the rex command-line utility. It provides specific instructions for the agent to execute shell commands such as rex __internal skill, rex whoami --json, and rex --help to manage terminal sessions, tabs, and panes.
[INDIRECT_PROMPT_INJECTION]: The skill implements a pattern where the agent ingests and follows instructions generated at runtime by the output of a local tool. While this represents an attack surface, it is a standard mechanism for CLI-integrated skills.
Ingestion points: The output of the rex __internal skill command serves as the primary source of instructions for the agent (referenced in SKILL.md).
Boundary markers: No explicit delimiters or instructions to ignore potentially malicious embedded instructions are present in the provided markdown.
Capability inventory: The skill description indicates capabilities for running commands, inspecting terminal output, and controlling connected Rex applications.
Sanitization: No sanitization or validation of the command output is performed within the skill instructions.

rex