Atlas › 01 Agent Skills for Browser Automation › Architecture Decision Records for an AI Test Automation Framework☰ Read as one page
Architecture Decision Records for an AI Test Automation Framework
10.1ADR-001: Agent-Driven vs Traditional Test ExecutionTraditional test frameworks (Selenium, Playwright, Cypress) execute tests as deterministic scripts. AI agents introduce non-deterministic…10.2ADR-002: Playwright CLI Skill as Primary Browser InterfaceWe evaluated three approaches: 1. Playwright MCP server (structured tools, streamed accessibility trees) 2. Playwright CLI taught via…10.3ADR-003: Test Definition FormatTests need to be defined in a format the agent can read and execute. Options: 1. Natural language descriptions 2. Gherkin (Given/When/Then)…10.4ADR-004: Fresh Sessions for CI, Reused Named Sessions for DevelopmentThe Playwright CLI manages browser sessions in the .playwright-cli/ workspace. Named sessions (-s=name) persist across commands; a session…10.5ADR-005: Snapshot-and-Screenshot DebuggingWhen a test fails, the agent needs to understand what went wrong. Options: 1. HTML dump of the page 2. Screenshot capture 3. Accessibility…10.6ADR-006: Error Recovery StrategyWhen a command fails (stale ref, timeout, unexpected state), the agent needs a strategy.10.7ADR-007: Test Result ReportingTest results need to be consumable by: 1. The AI agent (for reasoning about pass/fail) 2. Developers (for debugging) 3. CI systems (for…10.8ADR-008: Parallel Test ExecutionRunning tests sequentially is slow. But browser tests have shared state issues.