agentic-browser-testing

What problem does it solve?

This Skill replaces brittle, selector-heavy browser tests with a goal-driven agent that can explore changing UIs through the accessibility tree and prove the final outcome with an explicit oracle.

Core Features & Use Cases

• Intent-driven E2E testing: Validate fast-changing workflows without maintaining hardcoded click paths.
• Deterministic agent runs: Pin the model, set temperature to zero, seed state, and enforce a step budget so results are repeatable.
• Snapshot-first interaction: Use accessibility snapshots, role-based refs, and structured assertions instead of screenshots or pixel coordinates.
• CI gating and graduation: Block merges on machine-readable pass or fail results, then promote stable flows into scripted Playwright tests.
• Use case: A QA team can smoke-test a churning checkout or dashboard flow, catch false passes, and later turn the stable path into a durable regression test.

Quick Start

Ask the agent to run a goal-driven browser test against a seeded entry URL and require a final snapshot-based pass or fail verdict.