linux-at-spi2_skill

Overview

This skill provides expert guidance and patterns for safe AT-SPI2 automation on Linux desktops. It focuses on automating GTK and Qt applications via the D-Bus accessibility interface while emphasizing security, performance, and reliability. It is a high-risk capability and includes mandatory controls for system-wide access and input injection.

How this skill works

It inspects the AT-SPI2 accessible object tree and uses session D-Bus calls to query roles, states, text, and actions. The skill shows secure wrappers around pyatspi2/gi.repository.Atspi, event listener registration, and element discovery with timeouts and caching. Every operation is validated against blocklists, permission tiers, and audit logging to prevent sensitive interactions.

When to use it

• Automating UI flows in desktop applications that expose AT-SPI2 accessible objects.
• Building assistive-technology integrations that must query UI roles, states, or perform allowed actions.
• Creating test automation that interacts with UI elements where native APIs are unavailable.
• Monitoring UI focus and window events while avoiding high-volume event subscriptions.
• Any automation that requires strict audit trails, permission tiers, and safety checks.

Best practices

• Enforce permission tiers (read-only, standard, elevated) and block sensitive apps and roles such as password fields.
• Validate application identity via the desktop registry and maintain an app blocklist before interacting.
• Apply timeouts, rate limiting, and retry policies for all D-Bus calls to avoid hangs and flooding.
• Cache frequently accessed nodes with a short TTL and limit search scope to a specific application or role.
• Register only required AT-SPI2 events (avoid wildcard subscriptions) and log every operation with correlation IDs.

Example use cases

• Safe extraction of non-sensitive UI labels and states for automated testing in a CI environment.
• Controlled UI activation for accessibility testing while preventing input into password fields and terminals.
• Event-driven workflows that respond to window activation and focus changes without registering all events.
• Automated UI audits that record actions to an audit log for compliance reviews.
• Asynchronous element discovery and interaction for heavy tree traversals off the main thread.