The Agent Skills Directory

[COMMAND_EXECUTION]: The skill defines command patterns for invoking the opencode CLI tool via a shell, using user-supplied prompt strings as arguments.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a bridge for external data to reach shell execution, creating a potential injection surface. Ingestion points: User-provided input passed into the "<concise prompt>" placeholders within the shell command examples in SKILL.md. Boundary markers: None are present in the command templates to delimit or sanitize input strings. Capability inventory: Execution of the opencode tool with varying agents and flags. Sanitization: The skill provides no built-in sanitization, relying instead on the host agent's native approval flow and logging.

opencode-runner