Skip to main content
What you’ll learn
  • How to run a Test Audit using your AI agent
  • Why Test Audit is important
  • What it fixes: weak assertions, flaky tests, and missing coverage
Test Audit produces a 0–100 health score for a Playwright suite, plus prioritized issues with file-level evidence and fix recommendations. Audits are generated by your AI agent through the TestDino MCP server. Test Audit tab showing score badge, audit metadata, Overview tab with Issues list and Quick Actions sidebar

Setup

Test Audit runs through the TestDino MCP Server. Once MCP is configured, every audit is a single prompt to your AI agent.
1

Set up the TestDino MCP server

Follow MCP Overview to install the server, generate a PAT, and configure your client.
Skip this step if MCP is already working in your client.
2

Open your repo and run the audit prompt

Open the Playwright repo in your IDE so the agent can read your test files. Then send one of these prompts:
ScopeSample Prompt
SuiteRun a TestDino test audit on the full suite.
FeatureRun a TestDino audit on the <feature name> feature.
Spec FileRun a TestDino audit on <path/to/spec-file>.
Test CaseRun a TestDino audit on the test case <test name or ID>.
Critical and High issues are always reported, even outside the chosen scope.
3

View the audit in TestDino

Open AI Insights → Test Audit in TestDino ↗. The new audit appears at the top of the history with the score, issues, and full report.

Reading the Report

Audit Score

Every audit returns a single 0–100 score.
BandScoreMeaning
Excellent85–100Strong validation, low flake risk, well-structured
Fair65–84Localized weaknesses; targeted fixes recommended
Poor0–64Critical gaps in validation, stability, or coverage
A suite with multiple Critical issues cannot reach the Excellent band even if other dimensions look healthy.

Overview Tab

The Overview tab splits the audit into two panels side by side: a prioritized Issues list on the left and a Quick Actions sidebar on the right.

Issues

Each Issue card collapses to a single row with the severity badge, title, category badge, and citation count. Expanding the card reveals three blocks:
  • Summary: what the issue is and why it matters.
  • Recommendation: a concrete fix, often a one-line change at a file:line reference.
  • Evidence: one or more file:line references, each with a short observation.
Expanded Issue card showing severity badge, title, category, citation count, Summary, Recommendation, and Evidence with file:line references

Quick Actions

A sidebar listing up to four lightweight improvements that fall outside the main Issues. Suggestions are free-text and do not carry severity, category, or evidence metadata. Quick Actions sidebar listing lightweight improvement suggestions with file:line references

Full Report Tab

The Full Report tab renders the complete markdown audit document. Use the Download button to save it as a .md file for pull-request reviews or wikis. Full Report tab with title row, Download button, Executive Summary, and Category Snapshot table
SectionContents
Executive Summary2–3 sentences: score, top finding, trend direction
Category SnapshotIssue counts grouped by category
Test CompositionBreakdown of tests by type (Full Flows, Interactions, Render Checks, Page Loads, Accessibility, Other)
Score BreakdownHow the score was computed across dimensions
Audit CoverageFolders scanned and bounded counts for Critical/High patterns
Findings by SeverityIssues grouped by severity
Critical & High Issue MapEach cluster, why it matters, strongest evidence
RecommendationsQuick Wins, Medium Effort, Deep Refactors

Categories and Severity

Issue Categories

Each finding is tagged with one of nine categories.
CategoryWhat It Flags
Surface-Level TestsTests only check page load or basic UI presence, not real behavior
Missing ValidationAn action runs but the important outcome is never asserted
Flaky or UnstableHardcoded waits, race conditions, shared state, order-dependent steps
Hard to MaintainBrittle selectors, repeated .first(), no fixtures or page objects
Missing ScenariosGaps in error, empty-state, mobile, accessibility, or modal coverage
Organization & OwnershipUnowned test.skip or test.fixme, weak tagging, quarantine bloat
Setup & ConfigurationRetries hiding flakes, weak CI artifacts, risky worker isolation
Duplication & OverlapMultiple weak variants that should collapse into one stronger test
General IssuesFindings that do not fit the categories above

Severity Levels

LevelDefinition
CriticalBroken product behavior can ship, or confidence in a major area is invalidated
HighWidespread reliability or validation weakness across multiple files or features
MediumImportant but localized issue
LowNarrow cleanup
A scope with more than 50% surface-level tests is automatically reported as Critical.

Audit History

Past audits are stored per project and shown in the picker at the top of the Test Audit tab. Audit history dropdown listing past audits with score badge, audit name, branch, timestamp, and pagination controls
  • Navigate: use the < and > arrows, or open the dropdown to jump to any audit.
  • Pagination: 10 audits per page.
  • Select: click any past audit to load its Overview and Full Report.
  • Delete: the trash icon removes the selected audit. This cannot be undone.
  • Re-run: ask your AI agent for a new audit at any time. New reports are added to history without overwriting previous ones.

Failure Analysis

Cross-run failure categorization and patterns

MCP Overview

Connect AI agents to TestDino

MCP Tools Reference

All MCP tool specifications

Test Run AI Insights

Per-run failure categorization

Test Case AI Insights

Per-case AI diagnosis

Project Settings

AI controls and access tokens