manual-testing_skill

Overview

This skill guides users step-by-step through manually testing whatever is currently being worked on, prioritizing automated verification and falling back to user checks only when necessary. It helps identify what can be tested automatically, runs those checks, and produces minimal, focused manual verification steps for anything Claude cannot verify. The goal is fast, reliable validation and clear next actions when issues appear.

How this skill works

The skill inspects recent changes, build/test outputs, logs, server status, file contents, and conversation context to identify testable behaviors. It runs all automatable checks (build, lint, unit/integration tests, API requests, DB queries, port checks) and reports results. For visual or experiential items it cannot automate, it generates concise sequential user verification steps with expected outcomes and selectable response options. Finally it summarizes automated passes/failures and user-verified results with recommended next steps.

When to use it

• After implementing a feature that needs verification before merging or deploying
• When a reviewer says “test this” or “verify it works”
• Before asking QA or product for acceptance testing
• After fixing a bug to confirm the fix in context
• When unsure whether behavior is covered by automated tests

Best practices

• Automate everything possible; only ask the user for checks Claude cannot perform
• Run automated checks first and stop to fix failures before manual steps
• Keep manual steps short, specific, and single-purpose
• Provide clear expected results and 2–3 common failure options plus an Other field
• Report pass/fail per verification and recommend concrete next actions

Example use cases

• New UI component: run build/lint/tests, then ask user to confirm visual styling and interaction
• API endpoint change: run tests and curl requests, then ask user to exercise the client UI if needed
• Bug fix: run unit/integration tests and logs, then ask user to reproduce the failing scenario if automated checks pass
• Deployment smoke test: verify server starts, endpoints respond, and ask user to validate critical UI paths
• Accessibility check: run automated validators, then provide screen-reader/manual steps for user verification