qa-run-deep

What problem does it solve?

It solves the problem of missing, slow, or inconsistent QA validation by executing QA-*.md checklists through Playwright while simultaneously enriching each step with cross-linked SRS, PRD, implementation code, business rules, git history, and prior QA runs to produce higher-confidence results.

Core Features & Use Cases

• Deep QA execution with visual judgment: Runs MCP Playwright step sequences and performs AI-based screenshot/expectation comparison for natural-language expected outcomes.
• Context-rich regression and risk prediction: Collects a broad document + code + git-log context (up to an enforced 600K-line guard) and predicts higher-risk steps by intersecting recent commits with historical flakiness.
• Code ↔ QA consistency and meta-QA: Detects mismatches between what QA expects and what code/SRS actually implements, and flags stale QA assumptions based on timestamps and related TODO/SRS changes.
• Evidence-first reporting: Saves run artifacts to a structured QA-RUN report path including step evidence and categorized insights with measured vs inferred confidence.

Quick Start

Run qa-run-deep against the QA checklist file QA-MyFeature.md using deep mode for screenshot-based validation and prioritized execution.