Operate macOS application UI through computer_use MCP and the Computer → App → Element SDK. Use when observing, clicking, or typing on a macOS MCP host, including a remote host. Select by target desktop, not the Agent machine.
85%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
Low
Low-risk findings worth noting
The skill exposes the agent to untrusted, user-generated content from public third-party sources, creating a risk of indirect prompt injection. This includes browsing arbitrary URLs, reading social media posts or forum comments, and analyzing content from unknown websites.
The required workflow runs `app.read()` and `computer.read()` on whatever macOS app the task names, so UI text and screenshots from outsider-authored content (for example, inbound email or chat messages in a named app) reach the agent without it selecting a specific item.
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.