The Agent Skills Directory

[COMMAND_EXECUTION]: The skill resolves and executes local binaries such as orca, orca-ide, or orca-dev based on the environment variables and operating system. It includes specific logic to select the correct executable and avoids using the generic orca command on Linux to prevent unintended interaction with the system screen reader.
[INDIRECT_PROMPT_INJECTION]: The skill bootstraps its functionality by fetching the latest operational guide through the ORCA skills get orca-cli command, which represents a runtime ingestion of instructions.
Ingestion points: The output generated by the local orca binary when executing skills get orca-cli.
Boundary markers: No explicit markers or delimiters are used to wrap the fetched content, as it is intended to replace or extend the current operational context.
Capability inventory: The skill possesses the ability to control terminals, manage worktrees, and operate an embedded browser.
Sanitization: The instructions fetched from the CLI are treated as authoritative and are not subjected to additional validation or sanitization before being processed by the agent.

orca-cli