Course Outline
Session 1: Reproducible methodology and tool map (2h)
-
Module 1: From personal flow to reproducible methodology
-
Explicit team contracts: Sustaining SDD flow at coordinated collective scale.
-
Specification lifecycle: Refinement, validation, and traceability under high pressure.
-
AI reproducibility assurance: Structural consistency (same input → same output) in multi-agent environments.
-
-
Module 2: Multi-tool operational map
-
Local context control: Configuration of environments and guide files (
CLAUDE.mdvs.copilot-instructions.mdvs.OpenCode). -
Instructional guidance structure: Equivalences and discrepancies between Rules, AI Contexts, system instructions, and memory.
-
Multi-agent synchronization: Maintaining a single functional source of truth with heterogeneous tools.
-
Session 2: Precision in artifacts and calibration (2h)
-
Module 3: Precision in artifacts: The differences that matter
-
Specification boundaries: Technical differences and agent behavior regarding Functional Specs vs. Technical Specs.
-
Task granularity and sizing: Impact of massive tasks (code breakage) vs. minuscule tasks (context loss).
-
Comparative response analysis: Course adjustments in Copilot, Claude, and OpenCode facing the same specification.
-
Production antipatterns: Live demonstration of recurrent failures and their impact on generated code.
-
-
Guided Practice: Working on real specifications (30 min)
-
Active analysis and real-time diagnosis of specifications provided by the team to detect ambiguities.
-
Session 3: Full lifecycle and change management (2h)
-
Module 4: Refinement, validation, and mid-flight changes
-
Iterative refinement techniques: Transition from diffuse needs to validated specifications without writing base code.
-
Operational feasibility checklist before delegating tasks to an agent.
-
On-the-fly modifications: Managing technical scope variations mid-cycle without corrupting previous tasks.
-
Structural deviation control: Procedures to regain control when the agent diverges from the goal.
-
-
Module 5: Multi-agent patterns and MCPs in corporate environments
-
Agentized paradigm limits: Formal criteria to avoid excessive structural complexity without value return.
-
Advanced development patterns: Hierarchical orchestration, parallel agents, and human control gates.
-
MCP Ecosystem (Model Context Protocol): Corporate implementation with Filesystem, Git, and Jira (Direct demonstration).
-
Session 4: Quality, security, and real case resolution (2h)
-
Module 6: Agent-based TDD and quality pipeline
-
Specification as test contract: Transformation of acceptance criteria into automated tests by AI.
-
Automation in PR and CI/CD: What to automate and which decisions to retain under strict human review.
-
Continuous integration differences: Coupling methodologies for GitHub Copilot, Claude, and OpenCode.
-
-
Module 7: Security, intellectual property, and real limits
-
Data security policies: Protection of sensitive information and advantages of Enterprise/Team plans.
-
Intellectual property: Regulatory scenario, authorship, and responsibilities in AI-assisted development.
-
Organizational guardrails: Ethical and internal frameworks to boost delivery without compromising technological assets.
-
-
Practical Resolution: Team real use cases (30 min)
-
Solution in team code: Direct work on the repository and issues provided by participants.
-
Requirements
-
Daily and consolidated practice in Software-Driven Development (SDD).
-
Prior experience and active use in the workflow of at least one of the following tools: GitHub Copilot, Claude, or OpenCode.
-
Solid knowledge in software architecture, specification management (functional and technical), and continuous integration flows (CI/CD).
-
Essential requirement: Willingness to provide real use cases and code examples from the team before the training starts.
Target Audience:
- A consolidated team of 20 professionals with daily SDD practice (Senior Developers, Tech Leads, and Software Architects who already use AI in their daily work and seek to optimize, standardize, and scale their workflow).
-
Essential requirement: Willingness to provide the trainer with real use cases and code examples from the team before training begins for a preliminary expectation-management meeting.
Testimonials (4)
What appears most useful is the fact that both theory and practical demonstrations are shared, which helps ground the knowledge and reveals its real-world utility.
Noelia Iglesias Martinez - Aplicaciones y Tratamientos de Sistemas S.A (knowmad mood)
Course - IA Generativa y SDD Avanzado
Machine Translated
the MCPs
Gustavo Lopez Garcia - Aplicaciones y Tratamientos de Sistemas S.A (knowmad mood)
Course - IA Generativa y SDD Avanzado
Machine Translated
the automation flow of the agents
Antonio Perez Caballero - Aplicaciones y Tratamientos de Sistemas S.A (knowmad mood)
Course - IA Generativa y SDD Avanzado
Machine Translated
AGENTS.MD concepts
Eduardo Palomares Medina - Aplicaciones y Tratamientos de Sistemas S.A (knowmad mood)
Course - IA Generativa y SDD Avanzado
Machine Translated
