computer-use
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill implements logic to resolve a local binary for execution. It checks the
ORCA_CLI_COMMANDenvironment variable,ORCA_DEV_REPO_ROOT, or searches the system PATH fororca-ideororca. The resolved executable is then used to run commands such asskills getandopen. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to load its actual instructions from the output of an external tool at runtime (
ORCA skills get computer-use). This makes the agent vulnerable to instructions returned by the command output rather than those contained within the skill file itself. - Ingestion points: Command output from
ORCA skills get computer-usein SKILL.md. - Boundary markers: None provided; the skill instructs the agent to load the guide and use it.
- Capability inventory: The skill utilizes CLI execution and provides instructions for driving a local GUI.
- Sanitization: None detected; the agent is directed to follow the commands and flags covered by the guide.
Audit Metadata