Spec-driven development isn't waterfall with markdown. It's not a productivity trap. It's also not a silver bullet. The honest version is more useful than either the evangelist or the cynic gets credit for — and it's why your AI coding agent stops shipping the wrong thing.
The pages below assume the prior ones were read. Complexity and audience size grow as you scroll: solo workflow first, team rollout in the middle, multi-repo and multi-agent at the end. Want a full spec to build an agent yourself? See the free agent specs library — every page is a kanban your coding agent can execute.
Hitting the prompting wall, hearing about SDD, deciding if it is hype.
Anatomy, sizing, acceptance criteria, mindset, substrate, lifecycle, longevity.
Orienting SDD against TDD, RFC, and PRD — practices you already know.
Landscape, multi-framework comparisons, head-to-heads, decision tree.
Agent stack, solo workflow, reviewing AI-generated PRs without a teammate.
Anti-patterns, drift, the honest math of AI-assisted coding.
Standardization, vertical slices, contracts, multi-repo coordination.
The 30-repo scaling cliff, how context travels between codebases.
Migrating existing codebases or applying SDD in specialized contexts.
The frontier of running Cursor + Codex + Claude Code in parallel.
Critique, career, frontier, resources, FAQ.
Connect your repo. Describe the problem. Get a spec your agents can execute.