The Agent Skills Directory

[COMMAND_EXECUTION]: The skill identifies and executes local binaries (orca, orca-dev, or orca-ide) based on the presence of environment variables or the development environment context.
Evidence: The instructions in SKILL.md direct the agent to resolve the executable path using ORCA_CLI_COMMAND or ORCA_DEV_REPO_ROOT and subsequently run subcommands like status, open, and terminal list.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to fetch its full operational guide dynamically by executing a command on the local system.
Ingestion points: The agent is instructed to run ORCA skills get orca-cli and process the resulting output as the primary source of instructions for managing worktrees, terminals, and the browser.
Boundary markers: Absent; there are no delimiters or instructions provided to the agent to treat the dynamic command output as potentially untrusted data.
Capability inventory: The orca binary provides capabilities for terminal interaction (read/wait/send), file system management via worktrees, and controlling an embedded browser.
Sanitization: Absent; the skill lacks any validation or filtering mechanisms for the dynamic content fetched from the CLI tool.

orca-cli