The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The instructions explicitly direct the agent to "re-run it with escalated permissions" if it encounters a runtime_access_denied error from the sandbox. This encourages the agent to bypass environment-imposed security boundaries.
[COMMAND_EXECUTION]: The skill requires executing various versions of an external binary (orca, orca-dev, orca-ide) determined by environment variables or local checkout state. The logic involves selecting and substituting executables at runtime, which can lead to executing unintended binaries if not carefully controlled.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a stub that fetches its core logic and version-matched guides at runtime using the ORCA skills get orchestration command. This creates a vulnerability surface where the output of the external tool defines the agent's subsequent behavior.
Ingestion points: Output from the ORCA skills get orchestration command in SKILL.md.
Boundary markers: None identified in the provided stub.
Capability inventory: Shell command execution, task dispatch, and orchestration management across agents.
Sanitization: No evidence of sanitization or validation of the fetched guide content before the agent acts upon it.

orchestration