The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using a binary resolved from environment variables such as ORCA_CLI_COMMAND or ORCA_DEV_REPO_ROOT. This resolution process determines which executable (orca, orca-dev, or orca-ide) is utilized for subsequent tasks.
[INDIRECT_PROMPT_INJECTION]: The agent is directed to load its primary instructions via ORCA skills get orca-cli. This discovery stub pattern means the agent's logic is fetched dynamically from the output of a local binary, creating a surface where the agent's instructions can be influenced by the tool's output.
Ingestion points: The agent reads instructions from the stdout of the ORCA command execution.
Boundary markers: The skill does not provide explicit delimiters or warnings to the agent to distinguish between the discovery stub and the dynamically fetched instructions.
Capability inventory: The agent is granted capabilities to operate terminals, worktrees, and an embedded browser based on these instructions.
Sanitization: There is no process described to validate or sanitize the instructions returned by the CLI tool.
[DYNAMIC_EXECUTION]: The executable path for the CLI is not fixed and is computed at runtime based on the presence of environment variables or specific directory structures, which can lead to the execution of different binaries depending on the system configuration.

orca-cli