The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill instructs the agent to "re-run it with escalated permissions" if a command fails with a runtime_access_denied error. This is a direct directive to bypass security sandbox boundaries or filesystem permissions that may be in place for the session.
[INDIRECT_PROMPT_INJECTION]: The skill uses the command ORCA skills get orca-cli to dynamically load a "version-matched guide" that dictates how the agent should interact with the tool.
Ingestion points: The output of the ORCA skills get orca-cli command (SKILL.md).
Boundary markers: Absent; the agent is instructed to load and follow the guide's content without specific delimiters or warnings to ignore embedded instructions.
Capability inventory: The skill claims capabilities including operating worktrees, terminals, repositories, and an embedded browser (SKILL.md).
Sanitization: Absent; there is no validation or filtering process for the content returned by the CLI tool before the agent adopts its instructions.
[COMMAND_EXECUTION]: The instructions provide a logic flow for the agent to resolve and execute various local binaries (orca-dev, orca-ide, orca) and environment variables (ORCA_CLI_COMMAND) to interact with the host system.

orca-cli