specverify

Verify your AI specs before execution.

The missing layer between writing specifications and handing them to coding agents.

The Problem

Every SDD tool writes specs. No SDD tool verifies them.

You use GSD, BMAD, Spec Kit, Superpowers — they all produce specifications. Then you hand those specs to a coding agent and hope for the best. When the agent produces wrong code, you tweak the prompt. But the problem isn't the agent. It's the spec.

What specverify Does

specverify adds a verification step between your spec-writing tool and your coding agent:

Your SDD Tool → Write Spec → specverify → Pass? → Execute with Agent

The core test: "Can a different AI session generate functionally equivalent code from this specification alone — without additional context?"

If yes, the spec passes. If not, specverify tells you what's missing.

Features (coming in v1.0)

• 13-item Spec Gate — Structural completeness verification
• Adversarial Review — Cross-model validation (different AI attacks your spec)
• Clarity Gate integration — Epistemic quality checks
• Tool-agnostic — Works with any SDD workflow
• Git-verified proof — Built and validated with two case studies (46 endpoints in 4.5 hours, hackathon battle test)

Status

🚧 Under active development. Full release planned for April 2026.

Follow progress: github.com/frmoretto/specverify

Background

specverify is the operationalized tool from the Stream Coding methodology's Spec Gate protocol. The concept was first presented at AI Engineer Europe 2026.

Author

Francesco Marinoni Moretto
LinkedIn · GitHub

License

MIT