The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires the agent to execute shell commands using the orca CLI or its variants (orca-ide, orca-dev). These commands are used to list applications, query window states, and perform desktop automation tasks.- [INDIRECT_PROMPT_INJECTION]: The skill is designed to read data from external, untrusted sources including accessibility trees, application window content, and visible UI text from tools like web browsers or chat clients. This creates a surface for instructions embedded in those applications to influence the agent.
Ingestion points: Outputs from orca computer list-apps, orca skills get computer-use, and accessibility tree inspections mentioned in SKILL.md.
Boundary markers: Not present; the instructions do not specify any delimiters or safety warnings to distinguish between control instructions and data read from the screen.
Capability inventory: The agent has the ability to execute shell commands via the orca binary and perform UI interactions like clicking, typing, and dragging.
Sanitization: There is no mention of sanitization, filtering, or validation of the text read from the desktop environment before it is processed by the agent.

computer-use