COURSE & TRAINING
Agentic Continuous Verification and Testing
Use AI agents across testing, CI/CD, and runtime verification while keeping evidence, guardrails, and human escalation in control.
Online
On request
Live public or private training with an instructor.
On-site
On request
On-site for public or team cohorts.
Private team cohort
On request
Private training tailored to your engineering team.
Who this training is for
QA and test automation engineers
Generate and repair tests with agents while preserving coverage, independence, and assertion strength.
DevOps and SRE engineers responsible for CI/CD or production reliability
Connect agentic analysis, evidence-based quality gates, and runtime verification to delivery workflows.
Platform engineers who design delivery guardrails and quality gates
Design scoped agent access, MCP integrations, audit trails, and deterministic escalation paths.
What you'll learn
Design the agentic verification lifecycle
Place generation, analysis, quality gates, runtime verification, and human escalation in one defensible workflow.
Generate and audit tests
Generate unit, API, and end-to-end tests from specifications, then review coverage, independence, and assertion strength.
Detect the trust trap
Distinguish a legitimate test repair from a weakened assertion that hides a regression.
Run evidence-based quality gates
Drive pass, fail, or escalate decisions from analysis reports and define deterministic fallbacks for agent failure.
Close the runtime feedback loop
Verify a deployed service against runtime signals and promote suitable findings into new test cases.
Bound security, cost, and latency
Apply least privilege, secret hygiene, injection controls, budgets, and audit trails to the agent loop.
PREREQUISITES
What participants should bring
Participants should be comfortable with Git, CI/CD pipeline authoring, Linux, containers, reading and modifying Python or JavaScript, and at least one test framework. Experience with an observability stack is helpful but not required. Prior experience with LLM agents, MCP, prompt engineering, or machine learning is not required.
PREPARATION
Before the course starts
Participants receive a pre-flight checklist for the lab environment. Bring a laptop that can run the selected development, container, test, and CI/CD tooling. The final stack is confirmed for each delivery.
CURRICULUM
Inside Agentic Continuous Verification and Testing
A public online cohort is planned for November 2026, with exact dates coordinated around registrations. Private cohorts are available online or on-site. The lab stack is confirmed for each delivery so current, supported tools can be used.
Foundations and agent mechanics
Day 1
Understand why static automation decays, how agents plan and use MCP tools, and how generation, analysis, and observability agents work together.
Foundations and agent mechanics
Day 1
Understand why static automation decays, how agents plan and use MCP tools, and how generation, analysis, and observability agents work together.
Theoretical Part
- Verification versus testing and the limits of agentic approaches
- Agent planning loops, tool selection, termination, and non-determinism
- MCP as an integration and security boundary
- Three-agent roles and handoffs across generation, analysis, and observability
- Human review, missing outputs, and downstream quality gate failures
Practical Part
- Baseline a test suite for runtime, flakiness, and maintenance effort
- Trace an agent through prompts, tool calls, decisions, and termination
- Run the full three-agent loop, break one handoff, and diagnose the effect on the quality gate
Generating, healing, and trusting
Day 2
Generate tests from specifications, inspect agent repairs, and constrain the security, cost, and latency of the agent loop.
Generating, healing, and trusting
Day 2
Generate tests from specifications, inspect agent repairs, and constrain the security, cost, and latency of the agent loop.
Theoretical Part
- Prompting patterns for unit, API, and end-to-end tests
- Test design for boundary values, error paths, and idempotency
- Repair taxonomy and assertion-strength guardrails
- Prompt injection, secret hygiene, least privilege, and auditability
- Token cost, latency budgets, and per-run controls
Practical Part
- Generate and compare test suites from the same specification
- Find a plausible repair that weakens an assertion and hides a regression
- Harden an MCP tool scope and instrument cost and latency
Pipeline, runtime, and rollout
Day 3
Wire agentic analysis into CI/CD, interrogate runtime verdicts, and leave with a trust model and rollout plan.
Pipeline, runtime, and rollout
Day 3
Wire agentic analysis into CI/CD, interrogate runtime verdicts, and leave with a trust model and rollout plan.
Theoretical Part
- Pass, fail, and escalate quality gate patterns
- Deterministic fallbacks for provider outages, timeouts, and conflicting verdicts
- Runtime verdicts grounded in metrics, logs, traces, and SLOs
- Promoting suitable runtime findings into test cases without uncontrolled suite growth
- Trust matrices based on blast radius and verdict confidence
- Metrics for defect escape, flakiness, maintenance effort, coverage drift, and cost
Practical Part
- Wire a report-driven gate and test its degradation path
- Trace an observability verdict to its queries, expose missing evidence, and promote one suitable finding into a test case
- Complete the capstone and draft a one-page team rollout plan
TECHNOLOGIES
Technology roles covered in Agentic Continuous Verification and Testing
The course is vendor-neutral. MCP is the integration boundary. The coding agent, test framework, analysis layer, and observability stack are selected for each delivery.
Need Agentic Continuous Verification and Testing for your team?
Run a private cohort online or on-site, adapted to your stack and learning goals.
COURSE INSTRUCTORS
Yassin Aziz Rekik
Lead Instructor - SRE & Observability
Associate Professor at HEPIA Geneva for more than 20 years and former EPFL researcher. He leads the Effective SRE training and brings deep practical experience in observability, SLO engineering, software delivery, and reliability practices.
Sonia Boussabeh
Instructor
Upcoming cohorts
Date / PeriodNovember 2026Online — Online via Teams
StatusPlanned
Actions
Frequently asked questions
Everything you need to know about Agentic Continuous Verification and Testing.