18. Spec Engineer

Evidence label: Capability. Spec-driven development is an observable practice, but current evidence does not establish Spec Engineer as a durable standalone title.

What they do

Write the structured specifications that AI agents use to plan, implement, and validate code. This capability bridges business intent, human review, and machine-executable acceptance criteria. It can belong to a product engineer, requirements analyst, QA engineer, technical writer, or domain expert.

Core responsibility

Writing specifications that satisfy three conditions simultaneously: precise enough for an agent to generate correct code from them, human-readable enough for a product manager to approve them, and stable enough to serve as the diff-able ground truth when the implementation drifts.

GitHub Spec Kit formalizes a Constitution, Specify, Plan, and Tasks workflow around versioned artifacts. Tool adoption does not establish Spec Engineer as a market title.

Required skills

Technical	Soft
Structured writing (Gherkin-style Given-When-Then or equivalent)	Precision under ambiguity
Understanding of agent failure modes on multi-file and long-horizon tasks	Negotiation across product, engineering, and domain constraints
Familiarity with SDD tools (Spec Kit, Kiro, Augment, Factory.ai)	Ability to distinguish what the spec must constrain vs what it should leave open
Version control discipline (specs versioned before code)

Entry paths

Technical writer with engineering background, QA engineer who understands requirements, product engineer frustrated by low signal-to-noise in AI outputs, business analyst moving into AI-adjacent work.