Workshops: Agentic Coding

Spec-Driven Development with AI Agents: Workshop

Spec-driven development reverses the usual way of working with AI: instead of improvising prompts, you first produce a precise specification of requirements, design and task plan, which the agent then implements step by step. This workshop shows experienced teams how to use the method for larger features and get more predictable results.

Who is this workshop for?

  • Experienced developers already using agents
  • Tech leads and software architects
  • Product owners working closely with engineering
  • Teams with larger features and several contributors
  • Teams hitting the limits of vibe coding

What you will take away

  • Write requirements that agents can implement unambiguously and verifiably.
  • Produce a technical design and task plan with the agent before any code is written.
  • Have tasks implemented incrementally and checked against acceptance criteria.
  • Keep specification and code in sync as requirements change.
  • Know tools and templates for spec-driven workflows and choose the right level of formality.

Agenda

  1. Module 1

    Why specifications

    • Limits of ad-hoc prompting on large features
    • The spec as a shared contract between human and agent
    • Overview of approaches and tools, e.g. plan modes, Spec Kit, Kiro-style workflows
  2. Module 2

    Writing requirements

    • User stories with testable acceptance criteria
    • Non-functional requirements and constraints
    • Exercise: a spec for a feature from your backlog
  3. Module 3

    Design and task plan

    • Developing and critically reviewing a technical design with the agent
    • Breaking work into small, independently verifiable tasks
    • Project constitution: principles, architecture rules, quality standards
  4. Module 4

    Implementation and verification

    • Implementing tasks with Claude Code, OpenAI Codex, Cursor or GitHub Copilot
    • Deriving tests from acceptance criteria
    • Subagents and parallel implementation
  5. Module 5

    Embedding it in the team

    • Versioning and reviewing specs
    • Propagating requirement changes cleanly
    • When the effort pays off and when it does not

From prompt to specification

For small tasks a good prompt is enough. For features spanning days, files and people, improvised prompts cause drift: the agent makes assumptions no one consciously made and the result diverges from what was wanted.

A specification makes those assumptions explicit and checkable. It doubles as documentation, as a basis for reviews and as context any agent or team member can read.

Where the method pays off

Spec-driven development is not an end in itself. We show how to match the level of formality to the task.

  • Larger features with multiple components
  • Regulated environments with documentation duties
  • Distributed teams and hand-offs
  • Projects for external clients

Format

For teams with a grounding in agentic work. Delivered by neooutput AI Solutions Architects and AI Code Engineers, remotely or on-site, in English or German.

Frequently asked questions

Isn't this just waterfall with AI?

No. The spec is lightweight, produced with the agent in a short time and revised iteratively. It does not replace agile ways of working; it makes individual features plannable for agents.

What tools do we need?

At its core, Markdown files in the repository and a coding agent with a plan mode. We present specialised tools, but they are not a prerequisite.

Can product owners attend?

Yes, and it is a good idea. The requirements part benefits greatly from their perspective, even though the implementation modules are technical.

Workshop

Get in touch

Want your agents to deliver larger features predictably?

Tell us a little about your team – we will get back to you within 24 hours with a proposal for content, format and dates.

Prefer to talk?

We only use your details to handle this request. See our privacy policy.