QE Agentic
Workflow Roadmap
To stay relevant in the age of AI, Quality Engineers must transition from manual execution to orchestration. Learn to act as an AI orchestrator and observer of AI outcomes—directing teams of specialised agents for day-to-day technical tasks. Every episode pairs a concept video with a hands-on runbook.
🎬 Series Overview & Overall Goal
This video introduces the Agentic Markdown SDLC (Software Development Life Cycle), a method designed to bridge the gap between high-level project management portals like Azure DevOps and the actual codebase (0:00 - 0:25).
Key Concepts
- Local Source of Truth: By treating the product backlog as a version-controlled codebase where epics, features, and user stories are stored as Markdown files, teams can enable native AI accessibility and automated requirement refinement (0:27 - 0:50).
- Dual-Purpose Files: Each requirement file contains YML front matter for routing and Markdown bodies with Gherkin scenarios to create executable specifications (0:52 - 1:03).
- Bidirectional Sync: A custom CLI tool synchronizes task statuses and hierarchies between local Markdown files and Azure DevOps, ensuring the portal remains updated without manual effort (1:05 - 1:16).
The End-to-End Cycle
- Pulling: Moving requirements into the IDE as version-controlled files (1:30).
- Auditing: Using AI to check requirements against INVEST criteria (1:45).
- Decomposition: Breaking down monolithic requirements into testable sub-stories (2:00).
- Syncing: Reflecting the hierarchy in Azure DevOps (2:15).
- Implementation: Leveraging AI to generate code and styles based on the context (2:30).
- Verification: Running automated E2E tests to confirm acceptance criteria (2:45).
Week 1 — Foundation (Days 1–7)
7 episodesQE as AI Orchestrator — Lab Playbook 1
Direct Gemini 3.5 Flash through a 4-phase local workspace setup, git acquisition, markdown task definition, and automated execution loop
Git Basics & Sandbox Isolation
Navigating repositories and isolating work branches for secure agent access
Requirements Architecture & The INVEST Audit
Auditing stories against the INVEST framework to ensure they are AI-ready
Decomposition & BDD Prompting
Breaking epics into bite-sized Given-When-Then criteria for multi-step agent execution
Context Engineering & State Persistence
Preserving agent context across terminal sessions and managing memory state files
Multi-Agent Delegation & Orchestration
Coordinating specialist subagents: researchers, code writers, and test verifiers
Closed-Loop Defect Tracking & Auto-Repair
Automated triage, reproducing failures, and letting agents submit verified fix branches
Week 2 — Enterprise CI/CD & Governance (Days 8–14)
7 episodesHeadless Agent CI/CD Runners & Secret Sandboxing
Containerizing agent CLI execution within GitHub Actions and Azure DevOps pipelines with zero-leak secret protection
Automated PR Review & BDD Criteria Auditing
Configuring agents to audit incoming PR diffs against Given-When-Then BDD specifications and highlight missing test cases
Policy-as-Code & Quality Gate Enforcers
Enforcing token spend caps ($0.50 budget), static code compliance, and build-time quality gates
Closed-Loop Self-Healing CI Test Runners
Intercepting failing CI test steps, parsing stack traces, and auto-submitting targeted patch PRs with regression test evidence
Human-in-the-Loop Approval & Interactive Routing
Routing high-impact agent PRs to Slack/Teams with explicit interactive reviewer approval hooks before merge execution
Enterprise Audit Ledgers & Compliance Tracking
Maintaining immutable JSON audit logs tracking agent decision trajectories, prompt hashes, and SOC2 compliance markers
Enterprise Autonomous QE Pipeline Capstone
Constructing an end-to-end enterprise CI/CD pipeline combining PR auditing, self-healing, token caps, and approval gates