My Skill: Input Framing & Boundary Definition

Objective

Prevent the AI agent from relying on pre-trained parametric memory of generic login screens (e.g., assuming social login, CAPTCHA, or specific password rules).

Implementation Rules

• Explicit Element Declarations: Feed the LLM a strict manifest of present UI selectors before requesting test scenarios.
• Scope Exclusion Declarations: Explicitly list forbidden features that do not exist on the specific VWO login variant under test.

Applied Prompt Pattern

VWO CONTEXT

• App Name: Visual Website Optimizer (VWO)
• Module: Authentication / Login Screen
• Base URL: https://app.vwo.com/login

STRICT INVENTORY

The VWO login screen consists ONLY of:

1. Email Input (id="login-email")
2. Password Input (id="login-password")
3. Login Button (id="login-submit")
4. "Forgot Password?" Link (class="forgot-pwd")
5. Single Sign-On (SSO) Link (id="sso-login")

FORBIDDEN ASSUMPTIONS

DO NOT generate test cases referencing:

• Social Auth (Google, Slack, Microsoft)
• CAPTCHA / reCAPTCHA
• "Remember Me" checkbox
• "Show Password" eye icon/toggle
• Arbitrary account lockout limits (e.g., "Locked out after 3 attempts")