The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is centered on shell interaction via the orca binary or its variants. It provides logic for the agent to resolve the correct executable path using environment variables such as ORCA_CLI_COMMAND and ORCA_DEV_REPO_ROOT, or fallback defaults like orca-ide depending on the operating system.
[PROMPT_INJECTION]: The skill instructs the agent to retrieve its primary operational guide at runtime by executing ORCA skills get orca-cli. This design creates a surface for indirect prompt injection, as the agent's logic is influenced by external tool output rather than static instructions.
Ingestion points: The agent is explicitly told to read the output of a shell command (skills get) to determine subcommands, flags, and workflow details.
Boundary markers: There are no specified delimiters or 'ignore embedded instructions' warnings provided to distinguish the fetched content from the agent's core instructions.
Capability inventory: The orca binary possesses significant capabilities, including the ability to manage worktrees, repositories, terminals, and an embedded browser, as well as the ability to 'spawn' or 'handover' tasks to other agents.
Sanitization: The instructions provide no mechanisms for validating, filtering, or sanitizing the tool's output before the agent adopts it as functional guidance.

orca-cli