AI CompendiumA public chart of the language
Home / Patterns of Work / Spec-Driven Development
Patterns of Work

Spec-Driven Development

Write the specification first; let the implementation follow.

Definition

A workflow where a precise written spec — behaviour, constraints, acceptance criteria — precedes and guides implementation, whether by a person or an agent. The spec is the reviewable artifact.

It suits agentic work because agents execute well-specified tasks far more reliably than they infer intent. Reviewing a spec is also cheaper than reviewing a large diff.

The spec is the reviewable artefact that survives; the code is disposable and increasingly regenerable. That inverts the usual review target — a productive week might be one page of precise specification and a great deal of generated implementation.

How people use it
Standardising how the team works with agents.
"No agent starts without a spec in the ticket. Review the spec, then let it implement."
Worked example

Reviewing the spec, not the patch

A feature is about to be implemented by an agent.

Get the behaviour, constraints and acceptance criteria agreed in writing first. Corrections at that stage cost minutes; corrections after implementation cost a rewrite, and the spec is what lets anyone else judge the result without reading every line.

Sources
  • docsBest practices for using GitHub Copilot to work on tasks ↗GitHub docs · 2026

    Recommends bounded tasks with explicit acceptance criteria, and names what is a bad fit for unattended delegation: ambiguous, security-sensitive, production-critical or deeply domain-specific work.

  • blogDemystifying evals for AI agents ↗Anthropic Engineering · 2026

    Gives working definitions for harness, task, trial, grader, transcript and outcome, and argues you must evaluate the model and harness together and judge final environment state, not just the last message.

Related terms