The Agent Skills Directory

[PROMPT_INJECTION]: The skill facilitates indirect prompt injection by design. It enables the flow of instructions and data between multiple agents, where one agent's output can be injected into another agent's context.
Ingestion points: Messages from other agents are received via orca orchestration check and terminal output is read via orca terminal read (SKILL.md).
Boundary markers: The skill does not define explicit boundary markers or instructions to ignore embedded commands within the passed data.
Capability inventory: The skill can execute commands via orca terminal create --command, send text to shells via orca terminal send, and inject preambles into other agents via orca orchestration dispatch --inject (SKILL.md).
Sanitization: No sanitization or validation of the content being passed between agents is described in the command surface.
[COMMAND_EXECUTION]: The terminal management interface allows for arbitrary command execution.
Evidence: The command orca terminal create --command <cmd> allows launching any shell command in a new terminal. Additionally, orca terminal send --text <text> can be used to execute commands in existing terminal sessions (SKILL.md).

orchestration