Skip to main content
Test planning, test analysis, test design, and test implementation — Bugzy covers the full test creation lifecycle. It explores your application, understands its features, and writes comprehensive tests, acting as a QA teammate who handles the heavy lifting of building and maintaining test coverage. The process runs in two phases: first a structured test plan, then executable test cases with automated Playwright scripts.

​
Phase 1: Test plan

The generate-test-plan task takes a product description and produces a comprehensive test plan. Inputs:
  • Product description (from dashboard)
  • Live application (Bugzy explores with a real browser)
  • Existing project context and docs
Process:
1

Browser exploration

The browser-automation subagent opens your app and systematically navigates pages, forms, and flows.
2

Cross-reference with product description

Bugzy compares what it finds in the live app with the product description you provided.
3

Identify test scenarios

Happy paths, edge cases, error states, and boundary conditions are catalogued.
4

Pull product specs

If Notion or Jira is connected, the documentation-researcher subagent pulls product specs for additional context.
5

Generate test plan

A structured test plan is generated and committed to the repo.
Output: test-plan.md — a markdown document with test areas, scenarios, and expected behaviors. Test plan structure:

​
Phase 2: Test cases and automation

The generate-test-cases task converts the test plan into two artifact types.

​
Manual test case documents

Located in test-cases/TC-XXX.md, each document includes:
  • Preconditions
  • Numbered steps
  • Expected results
  • Test data

​
Automated Playwright scripts

Located in tests/specs/*.spec.ts, each script follows these conventions:
  • TypeScript with Page Object patterns
  • Proper waits and assertions
  • Environment variables for secrets
  • @smoke and @regression tags
1

Read test plan

The agent reads the test plan and identifies testable scenarios.
2

Generate Playwright files

The test-engineer subagent generates Playwright test files area by area.
3

Create Page Objects

Shared interactions are extracted into reusable Page Object classes.
4

Validate against live app

Tests are run against the live application to verify they pass.
5

Auto-fix failures

Failing tests are automatically fixed — selectors updated, timing adjusted.
6

Commit artifacts

All artifacts are committed to the repo.

​
Output artifacts

​
Browser exploration

The browser-automation subagent uses Playwright CLI to explore your application:
  • Navigates the base URL and crawls pages
  • Interacts with forms, buttons, and dropdowns
  • Takes screenshots of each page
  • Identifies dynamic content, auth walls, and multi-step flows
Exploration is bounded by the base URL domain — Bugzy does not navigate to external sites.

​
Knowledge base

During test generation, the agent builds a knowledge-base.md file that captures:
  • Application structure and navigation
  • Common selectors and interaction patterns
  • Authentication flows
  • Flaky areas and environment-specific behaviors
This knowledge base is consulted in future runs, so Bugzy’s understanding of your application improves over time.