Dugi Research Lab

Interactive framework / v0.1

GitOps Maturity Model

From versioned desired state to evidenced fleet operations: a cumulative Level 0–5 assessment.

Draft, author-defined framework. This is a structured self-assessment, not an audited rating, industry benchmark, or compliance determination. All answers stay in this page. Nothing is saved automatically.

Assess the evidence.

Answer for one defined team, service, or organization. Use the same scope across all five dimensions. “Unknown” scores zero because evidence is unavailable, not because failure is proven.

Is desired state declared and versioned with retained change history?

Does a controller pull desired state continuously and report or reconcile drift?

Are changes reviewed, identities scoped, and secrets kept outside plaintext Git?

Are reconciliation alerts, rollback or forward-fix, and disaster recovery regularly exercised?

Are policy, multi-cluster rollout boundaries, and measured improvement applied consistently?

How the model works.

Levels are cumulative. Each dimension earns its level only when it scores 3 and every preceding dimension also scores 3. A gap at Level 2 caps the result at Level 1 even if later controls are evidenced. Level 0 means the foundational evidence gate is unmet.

The five gates cover desired state, reconciliation, guarded changes, recovery, and fleet improvement. This progression is defined by this lab; it is not an OpenGitOps certification.

Background: OpenGitOps principles describe declarative, versioned, automatically pulled, continuously reconciled state. Our scoring and additional maturity gates are independent design choices.

Evidence rubric and limitations

0 = absent or unknown; 1 = planned or informal; 2 = implemented with incomplete or inconsistent evidence; 3 = implemented with repeatable evidence and review. Self-reported inputs are unverified. Five broad dimensions cannot capture every organizational risk. Use results to start a review, not to replace one.

Author & source

Dugi Selmanaj · AI-assisted draft, September 2026. Inspect the implementation and propose improvements ↗