Dexter — AI Remediation Engine

Dexter is the AI-powered remediation engine at the heart of the Cybrium platform. It takes findings from any scan type and generates, validates, and applies fixes — across cloud infrastructure, code repositories, and Kubernetes clusters.

Dexter is both the Cybrium mascot and the engine that closes the loop between "finding a vulnerability" and "fixing it."

Language	Rust
License	MIT
Source	github.com/cybrium-ai/dexter
Current version	0.1.0

How Dexter works

Finding discovered
    |
    v
Dexter analyzes the finding + resource context
    |
    v
Generates a fix (CLI command, Terraform, code patch)
    |
    v
Captures pre-state snapshot
    |
    v
[HITL gate] — human approves or rejects
    |
    v
Applies fix (dry-run or live)
    |
    v
Captures post-state snapshot
    |
    v
Verifies fix resolved the finding
    |
    v
Sends notification (email + platform)

Credential separation — PeriDex Engine

Dexter enforces strict credential scoping through the PeriDex Engine:

Scope	Access level	Used for
Scan	Read-only	Reconnaissance, vulnerability assessment, configuration reads
Fix	Write	Applying remediations, modifying resources, creating PRs

Scan credentials can never be used to apply fixes. Fix credentials require explicit approval through the Human-in-the-Loop gate before activation.

Supported providers

Provider	Fix methods
AWS	AWS CLI commands, CloudFormation, Terraform
Azure	Azure CLI commands, ARM templates, Terraform
GCP	gcloud commands, Terraform
Kubernetes	kubectl apply, Helm upgrades
Code	Git patch, pull request generation