Specification & Acceptance Control
Project Specification Compiler / SpecGate
Converted manually maintained project specifications into executable controls that validate mandatory sections, schema metadata, status vocabulary, bundle/source equality, registry IDs and repository drift.
Public technical artifactProblem
What this was built against.
AI-assisted implementation drifts when requirements, acceptance criteria and repository truth remain prose-only and manually maintained.
System
How the work closes the loop.
Project specification → deterministic schema/section validation → bundle equality → registry/path checks → CI report → correction before implementation/release.
Evidence now
What an employer can safely inspect.
- Public repository exposes an installable CLI (`specctl`) and pytest workflow.
- The compiler enforces eight mandatory sections, canonical metadata/status vocabulary and exact bundle/source equality.
- Related SpecGate work provides acceptance-criteria generation and deterministic output contracts.
Claim boundary
What this does not prove.
- Do not claim reduced rework or better delivery velocity until compared on real implementation cycles.
- The broader SpecGate/DriftGuard repositories overlap; this page treats specification compilation as the distinct capability.
Next evidence event
What would upgrade the proof.
Apply the compiler to one externally owned repository/specification and capture the concrete drift or ambiguity it catches before implementation.